About MiniMax H3 Max: Reference to Video
MiniMax H3 Max (Reference to Video) is a post-trained variant of MiniMax H3, tuned by fal for stronger prompt adherence and better aesthetics, and served on fal's inference stack. It generates video from a text prompt guided by multimodal references: up to 9 subject or style images, 3 motion video clips, and 3 audio clips.
It generates 5 to 15 second clips at 480P or 768P resolution with audio, including lip-synced dialogue, keeping subjects consistent with their reference images while following referenced motion and voices. References may total at most 12 files; video and audio clips run 2 to 15 seconds each with at most 15 seconds combined per type, and audio cannot be the only reference. Prompt expansion modes trade latency for prompt fidelity.
Ready to build with MiniMax H3 Max: Reference to Video?
Try MiniMax H3 Max: Reference to Video in the Workbench to prompt it, compare outputs, and iterate on prompts without writing any code. When you're ready to ship, call the same model from our API and build your own apps on top of it.
curl -sSf -X POST https://hub.oxen.ai/api/ai/videos/generate \
-H "Content-Type: application/json" \
-H "Authorization: Bearer $OXEN_API_KEY" \
-d '{
"model": "minimax-h3-max-reference-to-video",
"prompt": "@Image1 lifts off from a snowy branch and flies toward the camera in slow motion, snowflakes drifting through golden light, shallow depth of field, cinematic.",
"aspect_ratio": "adaptive",
"resolution": "768P",
"duration": 5,
"prompt_expansion_mode": "balanced",
"enable_safety_checker": true,
"sync_mode": false
}'