Text
Invent the scene from a chronological production brief.
Live MiniMax H3 partner API
Direct the shot.
Keep the references.
Generate MiniMax H3 video from text, first and last frames, or mixed image, video, and audio references. Picture and synchronized sound share one brief.
Model facts checked July 31, 2026SHOT DESK / 01Build a 2K MiniMax H3 clip
LiveMiniMax H3 · Text to Video
Text → VideoFunction
Text to videoCreate a complete shot from a written direction.
Describe one continuous shot in order: subject, action, camera, light, dialogue, ambience, and sound effects.
15 credits
Your video will appear here
02 / INPUT ROUTES
Three ways into one shot
Choose the route according to what must stay under your control—not according to how many files are available.
Prompt only
Keyframes
Anchor the opening frame and optionally the final composition.
Omni Reference
Carry visual identity, movement, voice, music, or material language into a new shot.
03 / MODEL EVIDENCE
Official MiniMax H3 outputs
All three examples below are linked directly from MiniMax’s current video-generation documentation.
Text to video
Prompt-led composition and motion.
First and last frame
Opening and ending frame control.
Omni reference
Mixed-reference visual and audio direction.
04 / CURRENT SPECIFICATION
A model page with the limits visible
MiniMax H3 is a general-purpose multimodal video model that creates picture and synchronized sound from one production brief. Start from text, animate a first frame with an optional last frame, or use mixed image, video, and audio references to carry characters, products, movement, voices, and visual language into a new shot.
- Output
- 2K video with optional synchronized audio
- Duration
- Official API: 4–15 seconds · Workspace: 5 / 10 / 15 seconds
- Keyframes
- First frame required; last frame optional
- Reference images
- Up to 9
- Reference videos
- Up to 3 · 2–15 seconds each · 15 seconds total
- Reference audio
- Up to 3 · requires an image or video reference
- Mixed limit
- Up to 12 reference files total
- Prompt
- Up to 7,000 characters in the official API; this workspace uses a conservative 5,000-character limit
05 / PRODUCTION LOOP
Validate one shot before scaling it
- 01
Pick the source mode
Start with text, a first frame, or the mixed reference workspace according to what must stay consistent.
- 02
Write one chronological shot
Order the subject action, camera movement, visual change, dialogue, ambience, and effects from first frame to last.
- 03
Check inputs and quote
Confirm duration, 2K output, framing, audio, and the live credit quote before generation.
- 04
Review picture and sound
Judge continuity, camera intent, sync, and the final frame before scaling the approved direction.
06 / CREDITS
Pay for approved production time
A 5-second 2K text or keyframe clip starts at 15 credits. Reference video length and extra images are included in the live quote.
Account
Create an account and add credits when you are ready.
$0forever
Credits purchased separately
- No subscription required
- All three MiniMax H3 modes
- Generation history
Creator
For regular short-form video work.
$19/ month
120 credits each month
- 120 credits every month
- Credits roll over
- Generation history
- Cancel anytime
Studio
For campaigns and production volume.
$49/ month
330 credits each month
- 330 credits every month
- Credits roll over
- Generation history
- Email support
Credit pack
Top up without a subscription.
$12once
50 credits
- 50 credits
- Never expires
- Works with any plan
07 / SOURCE LEDGER
What this page is based on
Primary model facts come from MiniMax. Partner pricing and endpoint availability come from the live inference provider.
01
MiniMax H3 guide
02Modes, duration, resolution, reference limits, formats, and API request shape.
MiniMax launch announcement
03Official release status, omni-reference positioning, and open-weights announcement.
Live partner endpoint
The MiniMax H3 partner route used by this generator for 2K inference.
08 / HELP
MiniMax H3 questions
Is MiniMax H3 available now?
Yes. MiniMax announced H3 on July 31, 2026, and this workspace uses the live MiniMax H3 partner endpoints for text, keyframe, and reference generation.
What can I use as a reference?
Reference mode accepts up to nine images, three videos, and three audio files, with no more than twelve files total. Audio references require at least one visual reference.
How long can a MiniMax H3 video be?
The official API supports 4 to 15 seconds. This workspace exposes 5, 10, and 15-second production presets at 2K.
Can MiniMax H3 generate sound?
Yes. Synchronized audio is enabled by default, and you can switch it off in the generator when the delivery needs a silent plate.
How are credits calculated?
The generator requests a server-verified quote for the selected duration and references before enabling generation. A 5-second 2K text or keyframe shot starts at 15 credits; reference media can increase the quote.
Is this the official MiniMax website?
No. This is an independent generation workspace using live partner infrastructure. Model specifications and the example videos are linked to MiniMax’s official documentation.