Nodes/Venice AI Nodes (unofficial)/Venice AI · Video Save Plus 🎬
ComfyUI Node

Venice AI · Video Save Plus 🎬

Saves MP4 with audio + workflow PNG/JSON to any folder. VIDEO, IMAGE+AUDIO, or passthrough of a finished MP4. Buttons: reveal, open, training copy, last frame, delete.

By Slartibart23·Created 25 days ago·Updated 24 days ago· 1
Venice AI · Video Save Plus 🎬
  • video
  • images
  • audio
  • filepath
◄filename_prefixvideo►
◄custom_output_path►
◄frame_rate24.00►
◄video_codech264►
◄crf19►
◄presetmedium►
◄pixel_formatyuv420p►
◄audio_bitrate192►
◄save_workflow_pngtrue►
◄save_workflow_jsonfalse►
◄png_compression4►
◄save_framenone►
◄frame_formatjpg►
◄frame_quality95►
◄copy_to_folder►
◄source_video_path►
◄prompt_text—►
◄seed0►
CategoryVenice AI/Save & Load

Inputs (21)

NameTypeDefaultDescription
filename_prefixSTRINGvideoBase filename. May contain subfolders (scene1/take) and placeholders: %date% (2026-08-21), %time% (143012), %seed% (from the seed input), %date:yyyy-MM-dd hh-mm-ss%. A 5-digit counter (_00001) is always appended.
custom_output_pathSTRINGAny folder on any drive. Created if missing. Empty = ComfyUI's default output folder.
frame_rateFLOAT24.001–120Only used for the 'images' input. With a VIDEO input or source_video_path the video's own frame rate is used.
video_codecCOMBOh264h264 = plays everywhere (Windows, web, phones). h265 = ~40% smaller at the same quality but not every player/browser supports it. Ignored in passthrough mode.
crfINT190–51Constant quality. Lower = better and larger. 0 = lossless, 17-19 = visually lossless, 23 = ffmpeg default, 28 = small. Main quality control. Ignored in passthrough mode.
presetCOMBOmediumEncoding speed vs. file size at the SAME quality (CRF). Does not change picture quality. Slower = smaller file, longer encode. Ignored in passthrough mode.
pixel_formatCOMBOyuv420pyuv420p = maximum compatibility (standard). yuv444p = full colour resolution, sharper edges/text, but many players and all browsers refuse to play it. Ignored in passthrough mode.
audio_bitrateINT19264–320AAC bitrate in kbit/s. 128 = speech, 192 = very good, 256-320 = music/transparent. In passthrough mode only used when 'audio' replaces the original track.
save_workflow_pngBOOLEANtrueSave the first frame as PNG with the complete workflow embedded. Drag & drop into ComfyUI restores the workflow.
save_workflow_jsonBOOLEANfalseAlso save the workflow as name.json (plain JSON, loadable via Workflow > Open; readable without image tools).
png_compressionINT40–9PNG is ALWAYS lossless. This only trades file size against write time: 0 = fastest/largest, 9 = smallest/slowest. 4 is a good balance.
save_frameCOMBOnoneAutomatically save frames as images next to the video (name_first Frame / name_last Frame / name_frame_00001). The 'Save Last Frame' button works independently of this. 'all' needs frames (images/VIDEO), not available for passthrough without frames.
frame_formatCOMBOjpgImage format for saved frames. jpg = small, lossy. png = lossless, larger (uses png_compression).
frame_qualityINT9510–100JPEG quality for saved frames (ignored for png). 90-95 = excellent, 100 = near lossless but large.
copy_to_folderSTRINGTarget folder used by the 'Save Training File' button. Copies mp4 + png + json + txt + frames with identical names.
videooptVIDEOVIDEO output of models/nodes with native audio (e.g. Grok Imagine Video, MiniMax). Frames, audio and frame rate are taken from it and RE-ENCODED with the settings above. A connected 'audio' input overrides its audio track.
imagesoptIMAGEFrames as IMAGE batch (used when no VIDEO is connected). In passthrough mode optional: only used for the PNG/frames.
audiooptAUDIOAudio track. Overrides the audio contained in a VIDEO input. In passthrough mode it REPLACES the file's own soundtrack (video stream is still copied, not re-encoded).
source_video_pathoptSTRINGPASSTHROUGH: path of a finished MP4 (e.g. the video_path output of Grok Imagine Video). The file is copied 1:1 to the destination - no re-encoding, no quality loss, audio kept - and renamed/numbered like everything else. Codec/CRF/preset/pixel_format are ignored. Leave unconnected to encode video/images as usual.
prompt_textoptSTRINGSaved as name.txt next to the video (training pair). Nothing is written if empty.
seedoptINT00–18446744073709550000Only used for the %seed% placeholder in filename_prefix.

Outputs (1)

NameTypeDescription
filepathSTRING—