> ## Documentation Index
> Fetch the complete documentation index at: https://docs.daydream.live/llms.txt
> Use this file to discover all available pages before exploring further.
# Update a Stream
> Updates pipeline parameters for an existing video processing stream
Rate limit: 300 requests per minute.
This endpoint allows you to update pipeline parameters for an existing stream. You can modify prompts, ControlNet settings, IP Adapter configuration, and other generation parameters.
## Request Body Structure
```json theme={null}
{
"pipeline": "streamdiffusion",
"params": {
"model_id": "Lykon/dreamshaper-8",
"prompt": "new prompt",
// ... only include parameters you want to change
}
}
```
The `pipeline` field defaults to `"streamdiffusion"` if omitted, so you can just send `params` for most updates.
## Hot-Swap vs Reload Parameters
The following parameters can be updated dynamically (no reload required):
* `prompt`, `negative_prompt`
* `guidance_scale`, `delta`
* `num_inference_steps`, `t_index_list`
* `seed`
* `controlnets.conditioning_scale`
* `ip_adapter` settings
* `ip_adapter_style_image_url`
All other parameters trigger a full pipeline reload (\~30s).
## Example: Update Prompt
```bash theme={null}
curl -X PATCH \
"https://api.daydream.live/v1/streams/${STREAM_ID}" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ${DAYDREAM_API_KEY}" \
-d '{
"params": {
"model_id": "Lykon/dreamshaper-8",
"prompt": "cyberpunk cityscape, neon lights"
}
}'
```
## Example: Adjust ControlNet Strength
```bash theme={null}
curl -X PATCH \
"https://api.daydream.live/v1/streams/${STREAM_ID}" \
-H "Content-Type: application/json" \
-H "Authorization: Bearer ${DAYDREAM_API_KEY}" \
-d '{
"params": {
"model_id": "Lykon/dreamshaper-8",
"controlnets": [
{
"model_id": "lllyasviel/control_v11f1p_sd15_depth",
"preprocessor": "depth_tensorrt",
"conditioning_scale": 0.7,
"enabled": true
}
]
}
}'
```
For full parameter documentation, see the [Parameters](/api/parameters/SD15) section.
## OpenAPI
````yaml PATCH /v1/streams/{id}
openapi: 3.1.0
info:
title: Daydream API Documentation
description: API documentation for Daydream API
version: 1.0.0
servers:
- url: https://api.daydream.live
description: Production server
security: []
paths:
/v1/streams/{id}:
patch:
tags:
- Streams
- public
summary: Update a stream
description: >-
Updates pipeline parameters for a specific stream by ID for the
authenticated user
operationId: updateStream
parameters:
- schema:
type: string
minLength: 1
in: path
name: id
required: true
description: ID of the stream to update
requestBody:
content:
application/json:
schema:
oneOf:
- $ref: '#/components/schemas/StreamDiffusion_4'
- $ref: '#/components/schemas/Scope_4'
discriminator:
propertyName: pipeline
mapping:
streamdiffusion:
$ref: '#/components/schemas/StreamDiffusion_4'
scope:
$ref: '#/components/schemas/Scope_4'
required: true
responses:
'200':
description: Default Response
content:
application/json:
schema:
oneOf:
- $ref: '#/components/schemas/StreamDiffusion_5'
- $ref: '#/components/schemas/Scope_5'
discriminator:
propertyName: pipeline
mapping:
streamdiffusion:
$ref: '#/components/schemas/StreamDiffusion_5'
scope:
$ref: '#/components/schemas/Scope_5'
'400':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'401':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'403':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'404':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'409':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'429':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
'500':
description: Default Response
content:
application/json:
schema:
type: object
properties:
success:
description: Always false for error responses
default: false
type: boolean
error:
description: User-facing error message
type: string
code:
description: >-
Stable machine-readable error code, e.g.,
STREAMS/NOT_FOUND
type: string
status:
description: HTTP status code
type: number
details:
description: Optional structured details about the error
requestId:
description: Request identifier for correlation
type: string
traceId:
description: OpenTelemetry trace id for correlation
type: string
required:
- success
- error
- code
- status
additionalProperties: false
security:
- Bearer: []
components:
schemas:
StreamDiffusion_4:
title: StreamDiffusion
type: object
properties:
pipeline:
type: string
enum:
- streamdiffusion
params:
oneOf:
- $ref: '#/components/schemas/SDTurbo_4'
- $ref: '#/components/schemas/SDXL_4'
- $ref: '#/components/schemas/SD1.5_4'
discriminator:
propertyName: model_id
mapping:
stabilityai/sd-turbo:
$ref: '#/components/schemas/SDTurbo_4'
stabilityai/sdxl-turbo:
$ref: '#/components/schemas/SDXL_4'
Lykon/dreamshaper-8:
$ref: '#/components/schemas/SD1.5_4'
prompthero/openjourney-v4:
$ref: '#/components/schemas/SD1.5_4'
required:
- pipeline
- params
Scope_4:
title: Scope
type: object
properties:
pipeline:
type: string
enum:
- scope
params:
description: Scope pipeline parameters. Accepts any JSON object.
type: object
additionalProperties: {}
required:
- pipeline
- params
StreamDiffusion_5:
title: StreamDiffusion
type: object
properties:
pipeline:
type: string
enum:
- streamdiffusion
params:
oneOf:
- $ref: '#/components/schemas/SDTurbo_5'
- $ref: '#/components/schemas/SDXL_5'
- $ref: '#/components/schemas/SD1.5_5'
discriminator:
propertyName: model_id
mapping:
stabilityai/sd-turbo:
$ref: '#/components/schemas/SDTurbo_5'
stabilityai/sdxl-turbo:
$ref: '#/components/schemas/SDXL_5'
Lykon/dreamshaper-8:
$ref: '#/components/schemas/SD1.5_5'
prompthero/openjourney-v4:
$ref: '#/components/schemas/SD1.5_5'
id:
description: Unique identifier for the stream
type: string
stream_key:
description: Unique key used for streaming to this endpoint
type: string
output_stream_url:
description: URL where the processed stream output can be accessed
type: string
created_at:
description: ISO timestamp when the stream was created
type: string
output_playback_id:
description: Playback ID for accessing the stream output
type: string
name:
description: Human-readable name of the stream
type: string
author:
description: ID of the user who created this stream
type: string
from_playground:
description: Whether this stream was created from the playground interface
type: boolean
gateway_host:
description: Gateway server hostname handling this stream
type: string
is_smoke_test:
description: Whether this is a smoke test stream
type: boolean
whip_url:
description: WebRTC WHIP URL for stream ingestion
type: string
required:
- pipeline
- params
- id
- stream_key
- created_at
- output_playback_id
- name
- author
- from_playground
- gateway_host
- is_smoke_test
- whip_url
additionalProperties: false
Scope_5:
title: Scope
type: object
properties:
pipeline:
type: string
enum:
- scope
params:
description: Scope pipeline parameters. Accepts any JSON object.
type: object
additionalProperties: {}
id:
description: Unique identifier for the stream
type: string
stream_key:
description: Unique key used for streaming to this endpoint
type: string
output_stream_url:
description: URL where the processed stream output can be accessed
type: string
created_at:
description: ISO timestamp when the stream was created
type: string
output_playback_id:
description: Playback ID for accessing the stream output
type: string
name:
description: Human-readable name of the stream
type: string
author:
description: ID of the user who created this stream
type: string
from_playground:
description: Whether this stream was created from the playground interface
type: boolean
gateway_host:
description: Gateway server hostname handling this stream
type: string
is_smoke_test:
description: Whether this is a smoke test stream
type: boolean
whip_url:
description: WebRTC WHIP URL for stream ingestion
type: string
required:
- pipeline
- params
- id
- stream_key
- created_at
- output_playback_id
- name
- author
- from_playground
- gateway_host
- is_smoke_test
- whip_url
additionalProperties: false
SDTurbo_4:
title: SDTurbo
type: object
properties:
model_id:
type: string
enum:
- stabilityai/sd-turbo
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- thibaud/controlnet-sd21-openpose-diffusers
- thibaud/controlnet-sd21-hed-diffusers
- thibaud/controlnet-sd21-canny-diffusers
- thibaud/controlnet-sd21-depth-diffusers
- thibaud/controlnet-sd21-color-diffusers
- daydreamlive/TemporalNet2-stable-diffusion-2-1
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: >-
Additional parameters for the preprocessor. For example, canny
edge detection uses 'low_threshold' and 'high_threshold'
values.
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
required:
- model_id
additionalProperties: false
SDXL_4:
title: SDXL
type: object
properties:
model_id:
type: string
enum:
- stabilityai/sdxl-turbo
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- xinsir/controlnet-depth-sdxl-1.0
- xinsir/controlnet-canny-sdxl-1.0
- xinsir/controlnet-tile-sdxl-1.0
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: Preprocessor parameters
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
ip_adapter:
description: >-
IP adapter — Turns on IP-Adapter style conditioning and is fully
hot-swappable. Available for SDXL, SDXL-faceid, SD1.5
type: object
properties:
type:
description: >-
Type of IP adapter. Use 'faceid' for SDXL-faceid models,
'regular' for others
default: regular
type: string
enum:
- regular
- faceid
enabled:
description: Whether IP adapter is enabled
type: boolean
scale:
description: Strength of IP adapter style conditioning
type: number
weight_type:
description: >-
Weight interpolation method for IP adapter style conditioning.
Controls how the style influence changes throughout the
generation process.
type: string
enum:
- linear
- ease in
- ease out
- ease in-out
- reverse in-out
- weak input
- weak output
- weak middle
- strong middle
- style transfer
- composition
- strong style transfer
- style and composition
- style transfer precise
- composition precise
ip_adapter_style_image_url:
description: >-
HTTPS URL or base64-encoded data URI (data:image/[type];base64,...)
of the style image to use. Base64 images are automatically cached
and uploaded to storage with content-based deduplication. Maximum
size: 5MB. When set, the runtime downloads the image and configures
the pipeline; if omitted, a default style image is used. For
'faceid' type, the image must contain a clear face. Available for
SDXL, SDXL-faceid, SD1.5
type: string
cached_attention:
description: >-
Cached attention (StreamV2V) configuration. Enable cached attention
to reuse key/value tensors across frames. ⚠️ NOTE: Enabling or
disabling cached_attention requires a pipeline reload as it uses a
different pipeline.
type: object
properties:
enabled:
description: >-
Whether this cached attention is active. Enable cached attention
to reuse key/value tensors across frames.
type: boolean
max_frames:
description: >-
Number of frames retained in the attention cache. Number of
historical K/V frames to retain. Limited by TensorRT engine
exports.
type: integer
minimum: 1
maximum: 4
interval:
description: >-
Cadence (number of frames) for refreshing cached key/value
tensors i.e How often (in number of frames) to refresh the
cache.
It is now **frame-based** (not seconds). It accepts integers 1-240, representing how many frames elapse between cache refreshes.
type: integer
minimum: 1
maximum: 240
required:
- enabled
- max_frames
- interval
required:
- model_id
SD1.5_4:
title: SD1.5
type: object
properties:
model_id:
description: Model to use for generation
type: string
enum:
- Lykon/dreamshaper-8
- prompthero/openjourney-v4
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
required:
- processors
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- lllyasviel/control_v11f1p_sd15_depth
- lllyasviel/control_v11f1e_sd15_tile
- lllyasviel/control_v11p_sd15_canny
- daydreamlive/TemporalNet2-stable-diffusion-v1-5
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: Preprocessor parameters
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
ip_adapter:
description: >-
IP adapter — Turns on IP-Adapter style conditioning and is fully
hot-swappable. Available for SDXL, SDXL-faceid, SD1.5
type: object
properties:
type:
description: >-
Type of IP adapter. Use 'faceid' for SDXL-faceid models,
'regular' for others
default: regular
type: string
enum:
- regular
enabled:
description: Whether IP adapter is enabled
type: boolean
scale:
description: Strength of IP adapter style conditioning
type: number
weight_type:
description: >-
Weight interpolation method for IP adapter style conditioning.
Controls how the style influence changes throughout the
generation process.
type: string
enum:
- linear
- ease in
- ease out
- ease in-out
- reverse in-out
- weak input
- weak output
- weak middle
- strong middle
- style transfer
- composition
- strong style transfer
- style and composition
- style transfer precise
- composition precise
ip_adapter_style_image_url:
description: >-
HTTPS URL or base64-encoded data URI (data:image/[type];base64,...)
of the style image to use. Base64 images are automatically cached
and uploaded to storage with content-based deduplication. Maximum
size: 5MB. When set, the runtime downloads the image and configures
the pipeline; if omitted, a default style image is used. For
'faceid' type, the image must contain a clear face. Available for
SDXL, SDXL-faceid, SD1.5
type: string
cached_attention:
description: >-
Cached attention (StreamV2V) configuration. Enable cached attention
to reuse key/value tensors across frames. ⚠️ NOTE: Enabling or
disabling cached_attention requires a pipeline reload as it uses a
different pipeline.
type: object
properties:
enabled:
description: >-
Whether this cached attention is active. Enable cached attention
to reuse key/value tensors across frames.
type: boolean
max_frames:
description: >-
Number of frames retained in the attention cache. Number of
historical K/V frames to retain. Limited by TensorRT engine
exports.
type: integer
minimum: 1
maximum: 4
interval:
description: >-
Cadence (number of frames) for refreshing cached key/value
tensors i.e How often (in number of frames) to refresh the
cache.
It is now **frame-based** (not seconds). It accepts integers 1-240, representing how many frames elapse between cache refreshes.
type: integer
minimum: 1
maximum: 240
required:
- enabled
- max_frames
- interval
required:
- model_id
SDTurbo_5:
title: SDTurbo
type: object
properties:
model_id:
type: string
enum:
- stabilityai/sd-turbo
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- thibaud/controlnet-sd21-openpose-diffusers
- thibaud/controlnet-sd21-hed-diffusers
- thibaud/controlnet-sd21-canny-diffusers
- thibaud/controlnet-sd21-depth-diffusers
- thibaud/controlnet-sd21-color-diffusers
- daydreamlive/TemporalNet2-stable-diffusion-2-1
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: >-
Additional parameters for the preprocessor. For example, canny
edge detection uses 'low_threshold' and 'high_threshold'
values.
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
additionalProperties: false
required:
- model_id
additionalProperties: false
SDXL_5:
title: SDXL
type: object
properties:
model_id:
type: string
enum:
- stabilityai/sdxl-turbo
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- xinsir/controlnet-depth-sdxl-1.0
- xinsir/controlnet-canny-sdxl-1.0
- xinsir/controlnet-tile-sdxl-1.0
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: Preprocessor parameters
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
additionalProperties: false
ip_adapter:
description: >-
IP adapter — Turns on IP-Adapter style conditioning and is fully
hot-swappable. Available for SDXL, SDXL-faceid, SD1.5
type: object
properties:
type:
description: >-
Type of IP adapter. Use 'faceid' for SDXL-faceid models,
'regular' for others
default: regular
type: string
enum:
- regular
- faceid
enabled:
description: Whether IP adapter is enabled
type: boolean
scale:
description: Strength of IP adapter style conditioning
type: number
weight_type:
description: >-
Weight interpolation method for IP adapter style conditioning.
Controls how the style influence changes throughout the
generation process.
type: string
enum:
- linear
- ease in
- ease out
- ease in-out
- reverse in-out
- weak input
- weak output
- weak middle
- strong middle
- style transfer
- composition
- strong style transfer
- style and composition
- style transfer precise
- composition precise
required:
- type
additionalProperties: false
ip_adapter_style_image_url:
description: >-
HTTPS URL or base64-encoded data URI (data:image/[type];base64,...)
of the style image to use. Base64 images are automatically cached
and uploaded to storage with content-based deduplication. Maximum
size: 5MB. When set, the runtime downloads the image and configures
the pipeline; if omitted, a default style image is used. For
'faceid' type, the image must contain a clear face. Available for
SDXL, SDXL-faceid, SD1.5
type: string
cached_attention:
description: >-
Cached attention (StreamV2V) configuration. Enable cached attention
to reuse key/value tensors across frames. ⚠️ NOTE: Enabling or
disabling cached_attention requires a pipeline reload as it uses a
different pipeline.
type: object
properties:
enabled:
description: >-
Whether this cached attention is active. Enable cached attention
to reuse key/value tensors across frames.
type: boolean
max_frames:
description: >-
Number of frames retained in the attention cache. Number of
historical K/V frames to retain. Limited by TensorRT engine
exports.
type: integer
minimum: 1
maximum: 4
interval:
description: >-
Cadence (number of frames) for refreshing cached key/value
tensors i.e How often (in number of frames) to refresh the
cache.
It is now **frame-based** (not seconds). It accepts integers 1-240, representing how many frames elapse between cache refreshes.
type: integer
minimum: 1
maximum: 240
required:
- enabled
- max_frames
- interval
additionalProperties: false
required:
- model_id
additionalProperties: false
SD1.5_5:
title: SD1.5
type: object
properties:
model_id:
description: Model to use for generation
type: string
enum:
- Lykon/dreamshaper-8
- prompthero/openjourney-v4
prompt:
description: >-
Text prompt describing the desired image. Can be a single string or
weighted list of (prompt, weight) tuples.
anyOf:
- type: string
- type: array
items:
type: array
prefixItems:
- type: string
minLength: 1
- type: number
prompt_interpolation_method:
description: >-
Method for interpolating between multiple prompts. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
normalize_prompt_weights:
description: >-
Whether to normalize prompt weights to sum to 1.0 for consistent
generation.
type: boolean
normalize_seed_weights:
description: >-
Whether to normalize seed weights to sum to 1.0 for consistent
generation.
type: boolean
negative_prompt:
description: Text describing what to avoid in the generated image.
type: string
guidance_scale:
description: >-
Strength of prompt adherence. Higher values make the model follow
the prompt more strictly.
type: number
delta:
description: >-
Delta sets per-frame denoising progress: lower delta means steadier,
less flicker but slower/softer; higher delta means faster, sharper
but more flicker/artifacts (often reduce CFG).
type: number
num_inference_steps:
description: >-
Builds the full denoising schedule (the 'grid' of possible
refinement steps). Changing it changes what each step number
(t_index_list value) means. Keep it fixed for a session and only
adjust if you're deliberately redefining the schedule; if you do,
proportionally remap your t_index_list. Typical range 10-100 with
default being 50.
type: integer
minimum: 1
maximum: 100
t_index_list:
description: >-
The ordered list of step indices from the num_inference_steps
schedule to execute per frame. Each index is one model pass, so
latency scales with the list length. Higher indices (e.g., 40–49 on
a 50-step grid) mainly polish and preserve structure (lower
flicker), while lower indices (<20) rewrite structure (more flicker,
creative). Values must be non-decreasing, and each between 0 and
num_inference_steps.
⚠️ NOTE: t_index_list must have 1–4 elements, non-decreasing, and
within [0, num_inference_steps]. The value should not go above 50.
minItems: 1
maxItems: 4
type: array
items:
type: integer
minimum: -9007199254740991
maximum: 9007199254740991
use_safety_checker:
description: Whether to use safety checker for content filtering
type: boolean
width:
description: >-
Output image width in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
height:
description: >-
Output image height in pixels. Must be divisible by 64 and between
384-1024.
type: integer
minimum: 384
maximum: 1024
lora_dict:
description: >-
Dictionary mapping LoRA model paths to their weights for fine-tuning
the base model.
type: object
nullable: true
use_lcm_lora:
description: Whether to use Latent Consistency Model LoRA for faster inference.
type: boolean
lcm_lora_id:
description: >-
Identifier for the LCM LoRA model to use. Example:
"latent-consistency/lcm-lora-sdv1-5"
type: string
acceleration:
description: >-
Acceleration method for inference. Options: "none", "xformers",
"tensorrt". TensorRT provides the best performance but requires
engine compilation.
type: string
enum:
- none
- xformers
- tensorrt
use_denoising_batch:
description: >-
Whether to process multiple denoising steps in a single batch for
efficiency.
type: boolean
do_add_noise:
description: >-
Whether to add noise to input frames before processing. Enabling
this slightly re-noises each frame to improve temporal stability,
reduce ghosting/texture sticking, and prevent drift; disabling can
yield sharper, lower-latency results but may increase flicker and
artifact accumulation over time.
type: boolean
seed:
description: >-
Random seed for generation. Can be a single integer or weighted list
of (seed, weight) tuples.
anyOf:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: array
items:
type: array
prefixItems:
- type: integer
minimum: -9007199254740991
maximum: 9007199254740991
- type: number
seed_interpolation_method:
description: >-
Method for interpolating between multiple seeds. Slerp provides
smoother transitions than linear.
type: string
enum:
- linear
- slerp
enable_similar_image_filter:
description: >-
Whether to skip frames that are too similar to the previous output
to reduce flicker.
type: boolean
similar_image_filter_threshold:
description: >-
Similarity threshold for the image filter. Higher values allow more
variation between frames.
type: number
minimum: 0
maximum: 1
similar_image_filter_max_skip_frame:
description: >-
Maximum number of consecutive frames that can be skipped by the
similarity filter.
type: integer
minimum: 0
maximum: 9007199254740991
skip_diffusion:
description: >-
Whether to skip the diffusion process. Any ControlNets or
diffusion-only parameters are ignored when enabled.
Example use cases:
- Stream the output of a preprocessor (e.g. live depth maps or pose skeletons),
- Run post-processors like RealESRGAN upscaler on externally provided frames,
- Warm a pipeline without paying the diffusion cost.
⚠️ NOTE: skip_diffusion is evaluated at pipeline creation time; switching mid-stream triggers a pipeline reload.
type: boolean
image_preprocessing:
description: List of image preprocessor configurations for image processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
image_postprocessing:
description: |-
List of image postprocessor configurations for image processing
⚠️ NOTE: realesrgan_trt processor requires a restart to change because it affects resolution.
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of image processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_preprocessing:
description: List of latent preprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
latent_postprocessing:
description: List of latent postprocessor configurations for latent processing
type: object
properties:
enabled:
description: Whether this processor config is active
type: boolean
processors:
description: List of latent processors to apply
type: array
items:
type: object
properties:
type:
type: string
enum:
- latent_feedback
enabled:
description: Whether this processor is active
type: boolean
params:
type: object
additionalProperties: {}
required:
- type
additionalProperties: false
required:
- processors
additionalProperties: false
controlnets:
description: >-
List of ControlNet configurations for guided generation. Each
ControlNet provides different types of conditioning (pose, edges,
depth, etc.). Dynamic updates limited to conditioning_scale changes
only; cannot add new ControlNets or change
model_id/preprocessor/params without reload.
type: array
items:
type: object
properties:
model_id:
description: >-
⚠️ NOTE: ControlNet model_ids must be unique. Additionally,
they must be compatible with the selected base model.
type: string
enum:
- lllyasviel/control_v11f1p_sd15_depth
- lllyasviel/control_v11f1e_sd15_tile
- lllyasviel/control_v11p_sd15_canny
- daydreamlive/TemporalNet2-stable-diffusion-v1-5
conditioning_scale:
description: >-
Strength of the ControlNet's influence on generation. Higher
values make the model follow the control signal more strictly.
Typical range 0.0-1.0, where 0.0 disables the control and 1.0
applies full control. Default: 1.0
type: number
minimum: 0
maximum: 1
preprocessor:
description: >-
Preprocessor to apply to input frames before feeding to the
ControlNet. Must be one of the supported preprocessors.
type: string
enum:
- blur
- canny
- depth
- depth_tensorrt
- external
- feedback
- hed
- lineart
- mediapipe_pose
- mediapipe_segmentation
- openpose
- passthrough
- pose_tensorrt
- realesrgan_trt
- sharpen
- soft_edge
- standard_lineart
- temporal_net_tensorrt
- upscale
preprocessor_params:
description: Preprocessor parameters
type: object
additionalProperties: {}
enabled:
description: >-
Whether this ControlNet is active. Disabled ControlNets are
not loaded. Default: true
type: boolean
control_guidance_start:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance begins. 0.0 means guidance starts from the beginning.
type: number
minimum: 0
maximum: 1
control_guidance_end:
description: >-
Fraction of the denoising process (0.0-1.0) when ControlNet
guidance ends. 1.0 means guidance continues until the end.
type: number
minimum: 0
maximum: 1
required:
- model_id
- conditioning_scale
- preprocessor
- enabled
additionalProperties: false
ip_adapter:
description: >-
IP adapter — Turns on IP-Adapter style conditioning and is fully
hot-swappable. Available for SDXL, SDXL-faceid, SD1.5
type: object
properties:
type:
description: >-
Type of IP adapter. Use 'faceid' for SDXL-faceid models,
'regular' for others
default: regular
type: string
enum:
- regular
enabled:
description: Whether IP adapter is enabled
type: boolean
scale:
description: Strength of IP adapter style conditioning
type: number
weight_type:
description: >-
Weight interpolation method for IP adapter style conditioning.
Controls how the style influence changes throughout the
generation process.
type: string
enum:
- linear
- ease in
- ease out
- ease in-out
- reverse in-out
- weak input
- weak output
- weak middle
- strong middle
- style transfer
- composition
- strong style transfer
- style and composition
- style transfer precise
- composition precise
required:
- type
additionalProperties: false
ip_adapter_style_image_url:
description: >-
HTTPS URL or base64-encoded data URI (data:image/[type];base64,...)
of the style image to use. Base64 images are automatically cached
and uploaded to storage with content-based deduplication. Maximum
size: 5MB. When set, the runtime downloads the image and configures
the pipeline; if omitted, a default style image is used. For
'faceid' type, the image must contain a clear face. Available for
SDXL, SDXL-faceid, SD1.5
type: string
cached_attention:
description: >-
Cached attention (StreamV2V) configuration. Enable cached attention
to reuse key/value tensors across frames. ⚠️ NOTE: Enabling or
disabling cached_attention requires a pipeline reload as it uses a
different pipeline.
type: object
properties:
enabled:
description: >-
Whether this cached attention is active. Enable cached attention
to reuse key/value tensors across frames.
type: boolean
max_frames:
description: >-
Number of frames retained in the attention cache. Number of
historical K/V frames to retain. Limited by TensorRT engine
exports.
type: integer
minimum: 1
maximum: 4
interval:
description: >-
Cadence (number of frames) for refreshing cached key/value
tensors i.e How often (in number of frames) to refresh the
cache.
It is now **frame-based** (not seconds). It accepts integers 1-240, representing how many frames elapse between cache refreshes.
type: integer
minimum: 1
maximum: 240
required:
- enabled
- max_frames
- interval
additionalProperties: false
required:
- model_id
additionalProperties: false
securitySchemes:
Bearer:
type: http
scheme: bearer
bearerFormat: JWT
````