MultiTrack Task Output

The MultiTrack editor gives you a timeline full of task segments, but ComfyUI needs a per-segment handle to actually generate anything. MultiTrack Task Output is that handle. Point it at the timeline, tell it which task segment you want, and it hands you that segment's user prompt, system prompt, task type, duration, and its media - the exact payload a video model node needs to generate that one clip. Loop over TASK_COUNT and you've turned the timeline into a driver for a whole multi-segment generation.

The inputs you actually set

• tracks_info - the timeline blob from MultiTrack Editor. Non-negotiable.
• task_index - which task segment to pull (zero-based). The number you'll increment in a loop.
• prompt_format - how the prompt string is assembled. The api default is what you want for feeding a model's conditioning node; other formats are there for different downstream consumers.

The optional images, audio, and video inputs matter only in slot-backed setups - when the timeline defers media loading (no slots referenced), this node materializes the media for the segment itself, so leave them disconnected and let it work.

What comes out

• SYSTEM_PROMPT and USER_PROMPT - the segment's prompts as strings. Wire these into the MultiTrack Prompt Enhancer, straight into a conditioning node, or into a text encoder.
• TYPE - the task type (t2v, i2v, r2v, v2v…), so downstream logic can branch.
• LENGTH - the segment's duration in frames.
• IMAGES, AUDIO, VIDEO - the media for this segment, as lists. For MiniMax-format timelines, segments with no audio/video correctly output None rather than an empty object - that was an actual bug fix in v1.2.1, so if you're on an old version, update.
• IMAGE_INDEXES - which images in the media list belong to this segment, as a string. Useful when you're passing a shared image list around and need to slice the right frames.

Why it's worth learning

Because generation models don't all speak the same dialect, the segment data is the bridge: t2v wants just the prompt, i2v wants the first image, v2v wants the whole video. This node's job is to make each segment's slice of reality available without you hand-copying values from the editor. Pair it with MultiTrack Info Output's TASK_COUNT and a loop, and one timeline produces N generations with no manual reshuffling between runs.

Install and gotchas

It's in ComfyUI-Easy-Media:

cd ComfyUI/custom_nodes
git clone https://github.com/yolain/ComfyUI-Easy-Media.git

restart, done (FFmpeg system-wide, as with everything in this pack). The traps are small but real: task_index out of range returns empty data, and if you connect your own images/audio/video inputs and the timeline also has media, you can end up double-feeding a downstream model. Keep the optional inputs disconnected unless you specifically need to override what the timeline has. And if prompts come back empty, check the editor - a task segment with no prompt set is a silent failure, not an error.