剪映合成视频非输出

JyComposeVideo (剪映合成视频非输出) takes the videos you've generated, arranges them into a single composed video at a canvas size and frame rate you choose, and hands back the output path and duration. The "剪映" in the name is Jianying - the Chinese version of ByteDance's CapCut editor - and the tell is in the inputs: it expects media, audio, effects, captions and track data as typed groups, which is the shape of a video-editing timeline rather than a bare list of files.

The "非输出" ("not an output") suffix is the part that matters when you're wiring. This variant is not marked as an output node, which means it sits in the middle of your graph without signaling "the workflow is done here." You'd use it when the composed video is an intermediate step - you want to preview it, run it through something else, or feed the path onward - rather than the final artifact. Its twin, JyComposeVideoOut, is the version you put at the end when the composition is the deliverable.

How it works

You feed it a group of media (the MEIDA_GROUP input), set the canvas (width, height), the frame rate (fps) and a name (draft_name), and it composes them into a timeline - with optional audio, effects, captions and an extra track layered in - then renders it out. It returns the resulting video's file path and its duration in seconds, so downstream nodes can record, display or chain off both. The custom types (MEIDA_GROUP, AUDIO_GROUP, EFFECT_GROUP, CAPTIONS_GROUP, TRACK) come from this same pack's video-tooling nodes - this node is the assembly step at the end of a pipeline that gathers media into groups first.

The inputs that matter

• medias (MEIDA_GROUP) - required; the video clips to compose. This is the one you must source from a compatible producer node.
• draft_name (STRING, default ComposeVideo) - the video's name. The author's own tooltip just says "视频名称" (video name); make it match the job so you can find the output.
• width / height (INT, default 1920×1080) - canvas resolution.
• fps (INT, default 30) - output frame rate.
• audios / effects / captions / track0 - optional layers for sound, transitions, subtitles and an extra track.

Outputs: 视频路径 (STRING - where the composed video landed) and 视频时长 (FLOAT - its length in seconds).

Install

Standard pack install, no extra model downloads for this node:

cd ComfyUI/custom_nodes
git clone https://github.com/yanlang0123/ComfyUI_Lam

and restart. Just be aware this is the China-centric corner of the pack - Jianying is CapCut's domestic sibling, and workflows using these compose nodes circulate mostly in Chinese-language shares (the pack's own README and UI are Chinese throughout).

Common issues

• "medias" input can't be filled - the MEIDA_GROUP type only accepts the pack's own media-group producer nodes. You can't drop a plain image or path in there.
• Nothing visibly saves - the composed file goes to a path returned as a string; grab the 视频路径 output and check that location rather than hunting the output folder.
• Durations don't add up - the 视频时长 output is the composed result's duration, which reflects how the timeline laid the clips out, not simply the sum of inputs.

If composition is the last thing your graph does, use JyComposeVideoOut instead - same behavior, but it correctly signals the end of the workflow to ComfyUI.