What does the output look like?
A reference-guided café scene
A woman speaks at an outdoor café in the H3-generated output shown in MiniMax’s launch article.
MiniMax official example: Introducing MiniMax H3What can MiniMax H3 do?
Open-weight native-audio video generation from text, first and last frames, or multimodal references at up to 4K
- Native-audio cinematic video
- Character and style consistency from reference images
- Motion and voice guidance from video and audio references
- Inputs
- Text, Image, Video, Audio
- Durations
- 5s, 6s, 7s, 8s, 9s, 10s, 11s, 12s, 13s, 14s, 15s
- Aspect ratios
- 21:9, 16:9, 4:3, 1:1, 3:4, 9:16
- Output options
- 480p, 768p, 2K, 4K
- Typical generation time
- About 10 minutes
- Reference mode accepts up to 12 files: 9 images, 3 videos, and 3 audio clips.
- Video and audio references must each be 2–15 seconds and audio needs a visual reference; 2K or 4K outputs use more credits than native 480p or 768p generation.
How much does MiniMax H3 cost?
Base estimate
$0.05–$0.16 USD / second
0.5–1.6 credits / second
Estimated dollar cost uses the Solo monthly plan at $0.10 USD per credit. Your cost per credit depends on your plan.
Example: a 5-second clip
$0.25–$0.80 USD(2.5–8 credits)
Reference media and video editing may add to this estimate.
Video cost scales with duration. Resolution, sound, reference media, and generation mode can affect the final charge. Check the total shown in Starpop before generating.
How to use MiniMax H3
- Click “Create with MiniMax H3” and sign in or create an account. After setup, Starpop opens a new video workspace tab with the model selected.
- Choose Frames for text or start/end images, or References for mixed image, video, and audio guidance.
- Describe the action and choose the duration and resolution. Attach a visual reference when using audio guidance.
- Review the credit estimate and generate. Check the result in your workspace, then download it or adjust your inputs to try again.