MiniMax H3 OW Fast 视频生成（5 合 1）

MiniMax H3 OW Fast is the speed tier of the H3 OW video line, and this node packs five of its modes into one dropdown: I2V Fast (single first frame), R2V Fast (up to 9 reference images), FL2VA and REF2VA Audio Drive Fast (first-frame/ref-image video driven by an audio track), and T2V Fast (plain text). The two Audio Drive modes are the reason this node exists - you hand it one image and one audio clip and it produces a video whose motion follows the audio. That's a "make this character talk to this voiceover" machine in a single node.

Everything here is the pack's standard API-wrapper arrangement: T8mars/ComfyUI_Seedance, key from api.seedance.nz, no local compute.

How it works

The node dynamically shows 0, 1 or 9 image slots depending on the mode, and only shows the audio input for the two Audio Drive models. I2V Fast strictly uses image1 and nothing else; R2V Fast uses image1..image9 (in connected order, compacted over gaps); both Audio Drive modes require exactly image1 plus one audio, which gets converted to WAV, uploaded, and sent to metadata.audio_urls as a single-element array. T2V Fast takes no media at all. Then it's the pack's usual upload → submit → poll → download loop.

The inputs that matter

• model - the five-way switch: minimax-h3-ow-i2v-fast, minimax-h3-ow-r2v-fast, minimax-h3-ow-fl2va-audio-drive-fast, minimax-h3-ow-ref2va-audio-drive-fast, minimax-h3-ow-t2v-fast.
• prompt - required for T2V and R2V; optional for I2V and the Audio Drive modes (though a hint helps the audio-driven ones).
• seconds - 5, 10 or 15.
• resolution - 480p or 720p. This is the "Fast" tier; don't expect 4K.
• ratio - one of the eight documented aspects (1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9, 21:9).
• image1..image9 and audio - per mode, as above.

Outputs: video, video_url, task_id, response.

Installing it

ComfyUI Manager → search "ComfyUI Seedance", or:

cd ComfyUI/custom_nodes
git clone https://github.com/T8mars/ComfyUI_Seedance.git
cd ../..
python -m pip install -r custom_nodes/ComfyUI_Seedance/requirements.txt

Restart, key from api.seedance.nz/console, Seedance API Config into api_config.

Common issues

• I2V Fast ignores your extra images - I2V Fast uses image1 only; the 9-slot expansion belongs to R2V.
• Audio Drive missing audio - both FL2VA and REF2VA require exactly one audio clip. It's required, not optional.
• Audio slot not showing - you're on I2V/R2V/T2V; only the two Audio Drive models expose audio.
• 480p looks soft - that's the Fast tier's contract. If you need sharper, the pack's standard Minimax_H3_OW_Video node (480p/720p too, but the full-quality tier) or a Hailuo H3 2K render is the upgrade.

For testing the audio-drive behavior, keep the clip short (5s) and the image simple. Once you've confirmed the motion tracks, that's when you spend on longer runs - Fast is meant for exactly that kind of cheap iteration.