Generate AI videos with native audio guided by reference images, videos, and audios using MiniMax H3 Open Reference to Video. 480p/768p and 5-15s clips.
MiniMax H3 Open Reference to Video generates coherent videos from prompts and multimodal references, guided by up to 9 reference images, 3 reference videos, and 3 reference audios with native stereo audio.
About this model
MiniMax H3 Open Reference to Video generates coherent 480p and 768p videos from text prompts and multimodal reference inputs. Guided by up to 9 reference images, 3 reference videos, and 3 reference audios, it synthesizes visual motion and native stereo audio in unison. Compare with MiniMax H3 Open Text to Video and MiniMax H3 Open Image to Video.
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $0.065/s (480p) or $0.13/s (768p) | Multimodal reference billing: $0.026 per image/audio and $0.065/s per reference video second. |
| Fal.ai | $0.20+ per second | Higher base per-second rates for reference-based video generation. |
| Replicate | Not available | Model not hosted natively on Replicate. |
Multimodal reference billing: $0.026 per image/audio and $0.065/s per reference video second.
Higher base per-second rates for reference-based video generation.
Model not hosted natively on Replicate.
** Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Prompt | string | Text description referencing <Picture 1..9>, <Video 1..3>, <Audio 1..3>. Native stereo audio is generated with the video. | A cinematic ocean wave at sunrise with <Picture 1> subject, highly detailed |
| Reference Images | array | Reference image URLs. Up to 9 images. | undefined |
| Reference Videos | array | Reference video URLs (up to 3, 480p). Total reference video duration up to 15s. | undefined |
| Reference Audio | array | Standalone reference audio URLs (up to 3, trimmed to 15s each). | undefined |
| Aspect Ratio | Enum (7 options) | - | 16:9 |
| Resolution | Enum (2 options) | - | 480p |
| Duration | Enum (11 options) | - | 5 |
Text description referencing <Picture 1..9>, <Video 1..3>, <Audio 1..3>. Native stereo audio is generated with the video.
A cinematic ocean wave at sunrise with <Picture 1> subject, highly detailedReference image URLs. Up to 9 images.
undefinedReference video URLs (up to 3, 480p). Total reference video duration up to 15s.
undefinedStandalone reference audio URLs (up to 3, trimmed to 15s each).
undefined-
16:9-
480p-
5Developer documentation
prompt referencing <Picture 1..9>, <Video 1..3>, or <Audio 1..3> along with your reference URLs.aspect_ratio (e.g., 16:9), resolution (480p or 768p), and duration (5 to 15 seconds)./api/v1/minimax-h3-open-reference-to-video using curl or your client library.curl -X POST "https://api.muapi.ai/api/v1/minimax-h3-open-reference-to-video" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"prompt": "A cinematic scene with <Picture 1> character walking, highly detailed",
"reference_images": ["https://example.com/char.jpg"],
"resolution": "480p",
"duration": 5
}'
Frequently asked
MiniMax H3 Open Reference to Video is a multimodal model that generates videos and native audio guided by up to 9 reference images, 3 reference videos, and 3 reference audios.
Use tags such as <Picture 1>..<Picture 9>, <Video 1>..<Video 3>, and <Audio 1>..<Audio 3> in your prompt text.
The model supports 480p and 768p resolutions with output durations ranging from 5 to 15 seconds.
Yes, native stereo audio is generated in sync with the visual motion and any reference audios provided.
Output video is billed at $0.065/sec (480p) or $0.13/sec (768p). Per-reference charges: $0.026 per image, $0.026 per audio track, and $0.065/sec per second of reference video duration.