MiniMax H3 Masking · Trim Source AV

H3 doesn't accept just any clip length. Its frame count has to satisfy length % 17 == 5 - valid runs like 5, 22, 39, 56, up to hundreds of frames - and if your source video is 143 frames you're stuck with a length that isn't valid. This tiny node fixes that by trimming your footage down to the largest valid H3 length, dropping trailing frames instead of padding or inventing any. It's a boring utility, and that's exactly why you want it: it removes a whole class of "why does my source refuse to load" errors.

What it does

Feed it frames (an IMAGE batch) and optionally the audio that belongs to it. It computes the largest valid 17k+5 frame count that fits inside your source, trims the video to that, and trims the audio to the matching duration at 24 fps so sound stays locked to picture. Frames are never resized or padded - if your source is 143 frames you get 124 back, not 143 with filler.

Outputs are the trimmed_frames, the trimmed_audio (an empty output if no audio was connected), h3_length with the exact valid count, and a human-readable trim_info string. That h3_length is the number to feed anywhere the workflow wants a raw H3 length.

It's a genuinely thin node. You'd typically put it right after loading source footage and before it hits the chain's masked AV or source-timeline path, so everything downstream sees a length H3 will accept.

Install and gotchas

Same pack install as the rest:

cd ComfyUI/custom_nodes
git clone https://github.com/ethanfel/ComfyUI-MiniMaxH3-Contex-Loop.git

Restart ComfyUI, or install via ComfyUI Manager under "MiniMax H3 Contex Loop". No extra pip dependencies; bring your own H3 weights (check the MiniMax H3 Community License - local weights are restricted in the US/EU/UK/Korea), and a current ComfyUI with native Add Guide support (PR #15439) helps.

Two things to keep in mind. First, this trims - if your footage is shorter than any valid run (under 5 frames), there's nothing it can return, so don't use it as a fix for too-short material. Second, the audio trim assumes 24 fps, which is H3's timeline; if your source is a different frame rate, resample it to 24 before connecting, or the audio duration won't match the picture.