ComfyUI Node
SilentComposite (render -> one always-silent video)
A ComfyUI node in OldTimeRadio/v2/video with 10 inputs and 2 outputs.
SilentComposite (render -> one always-silent video)
- silent_video_path
- report
◄base_video_path►
◄canvas_w1472►
◄canvas_h832►
◄fps25►
◄ffmpegffmpeg►
◄output_path►
◄gate_in►
◄clip_manifest_json{}►
◄upscale_engineoff►
◄upscale_devicecpu►
CategoryOldTimeRadio/v2/video
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| base_video_path | STRING | Base video to normalize (M1: the OTR_SignalLostVideo radio-floor mp4). Any audio is STRIPPED here (V-1). | |
| canvas_wopt | INT | 147216–7680 | — |
| canvas_hopt | INT | 83216–4320 | — |
| fpsopt | INT | 251–120 | — |
| ffmpegopt | STRING | ffmpeg | — |
| output_pathopt | STRING | Silent composite path. Empty -> <output>/otr/episodes/<stem>_silent.mp4. | |
| gate_inopt | STRING | Optional ordering signal (opaque STRING). | |
| clip_manifest_jsonopt | STRING | {} | OTR_VideoRenderBatch(mode=episode) clip manifest. When it carries per-beat clips the composite ASSEMBLES a frame-accurate CFR timeline (each beat conformed to its audio-derived frame count; gaps filled from base_video_path). Empty -> the single-base floor path. |
| upscale_engineopt | COMBO | off | Post-render upscale/enhance stage. Runs INSIDE the composite per-clip (sharpen=True branch only); floor / directory / black paths are unchanged. off = today's lanczos+unsharp, byte-identical. |
| upscale_deviceopt | STRING | cpu | cpu | cuda | cuda:N. Composite itself runs on CPU (ffmpeg); this widget selects the device the model loads on. Rejected invalid tokens fail loud. |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| silent_video_path | STRING | — |
| report | STRING | — |