ComfyUI Node
Difforum · Live Sampler
Realtime Difforum: queue once and watch it play inside the node. Pair with a 1-4 step model (SD-Turbo, SDXL-Turbo, LCM, DMD2) at ~512 px. `live_source` turns it into a magic mirror (webcam "0" or a video path); `stream_dir` / `spout_name` feed OBS, Resolume or TouchDesigner. Only the last `keep_frames` frames are returned, so long sessions do not fill RAM.
Difforum · Live Sampler
- model
- positive
- negative
- vae
- init_image
- direction
- options
- params
- camera
- strength
- prompts
- depth
- frames
- report
◄run_frames480►
◄steps2►
◄cfg1.2►
◄sampler_namelcm►
◄schedulersgm_uniform►
◄cadence1►
◄target_fps0.0►
◄energy0.50►
◄loop_cameratrue►
◄live_source►
◄source_blend0.90►
◄stream_dir►
◄spout_name►
◄keep_frames240►
◄live_previewtrue►
CategoryDifforum/4 · Render
Inputs (27)
| Name | Type | Default | Description |
|---|---|---|---|
| model | MODEL | The image model (SD1.5 / SDXL / Flux / turbo). MODEL output of your checkpoint or LoRA chain. | |
| positive | CONDITIONING | Positive conditioning (CLIP Text Encode). | |
| negative | CONDITIONING | Negative conditioning (CLIP Text Encode). | |
| vae | VAE | The VAE that matches the model. | |
| init_image | IMAGE | The first frame. Everything grows from this image. | |
| run_frames | INT | 4801–10000000 | How many frames to run (the loop wraps the camera). |
| steps | INT | 21–50 | Steps per frame. Turbo models: 1-4. |
| cfg | FLOAT | 1.20–15 | Prompt strength. Turbo / LCM / DMD2 models want 1-2; regular SDXL 4-7. |
| sampler_name | COMBO | lcm | Sampler. euler / dpmpp_2m for regular models, euler_ancestral or lcm for turbo. |
| scheduler | COMBO | sgm_uniform | Noise schedule. karras / normal for regular models, sgm_uniform for turbo. |
| cadence | INT | 11–12 | Diffuse every Nth frame for speed. |
| target_fps | FLOAT | 0.00–60 | Cap the speed. 0 = as fast as possible. |
| directionopt | DIFFORUM_DIRECTION | The Director's wire: camera, energy curve, scene prompts, keys and look in one cable. | |
| optionsopt | DIFFORUM_OPTIONS | Difforum · Render Options: manual colour, detail, symmetry, depth and chunk settings. Without it the Director's look decides. | |
| paramsopt | DIFFORUM_PARAMS | From Setup: frame count, fps, width and height. Plug the same Setup everywhere. | |
| cameraopt | DIFFORUM_CAMERA | A camera track from Camera (keys), Camera (expressions) or Camera Import. Overrides the Director's camera. | |
| strengthopt | DIFFORUM_SCHEDULE | A schedule for denoise per frame; replaces the Director's energy curve. | |
| promptsopt | DIFFORUM_PROMPT | Prompt travel from Prompt Travel (or the Director with CLIP connected). | |
| depthopt | IMAGE | A depth map (white = near), e.g. Depth Anything V2 on the first frame. Needed for real 3D parallax; without it 3D moves fall back to pseudo-3D. | |
| energyopt | FLOAT | 0.500–1 | Denoise per frame when no Director is connected. |
| loop_cameraopt | BOOLEAN | true | Repeat the camera when the run is longer than the clip. |
| live_sourceopt | STRING | '' = off, '0' = webcam, or a video path: mixed in before each frame. | |
| source_blendopt | FLOAT | 0.900–1 | How much of the live source enters each frame. |
| stream_diropt | STRING | Folder to write live PNG frames to (for another app). | |
| spout_nameopt | STRING | Spout sender name (Windows) for Resolume / TouchDesigner / OBS. | |
| keep_framesopt | INT | 2401–100000 | Frames kept in memory for the output (ring buffer). |
| live_previewopt | BOOLEAN | true | Show the frames inside the node while running. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| frames | IMAGE | — |
| report | STRING | — |