Seedance 2.5Ads & brand films
On-screen text and logos render accurately.

Turn a prompt, a first and last frame, or image, video and audio references into 4–15 second videos with native stereo sound. Output up to 2K in 21:9, 16:9, 1:1 or 9:16.
Halloween Special·Open your mystery box and claim your Halloween treat!
4–15 seconds · up to 2K · 24 fpsProduct shots, character scenes and dialogue clips generated with MiniMax H3, with native stereo audio.
MiniMax H3 is a general-purpose multimodal video model released by MiniMax on July 31, 2026. It reads text, images, video and audio together and generates video with native stereo sound in a single pass.
Describe the subject, action, camera move and sound. H3 follows complex instructions and renders on-screen text and brand names accurately, so prompts for ads and product films work as written.

Upload one image as the first or last frame, or two images for both, then describe the motion in between. Ideal for product reveals and planned transitions.

Combine up to 9 images, 3 video clips and 3 audio clips (12 files max, video and audio up to 15 seconds total) to lock characters, motion and voice. H3 can also transfer motion from one video to another.

Copy a tested MiniMax H3 prompt, swap the subject or setting, and generate. Add a line for sound or dialogue — H3 generates audio with the video.
Visual example

Reference prompt
Seedance 2.5 · Text to Video
An ancient Chinese city during a lantern festival at night. Fireworks bloom above tiled roofs, warm lanterns line the streets. The aerial camera descends slowly past the rooftops toward a lively market. Deep blue sky, warm practical light, an expansive cinematic atmosphere.
MiniMax positions H3 for advertising, branding, e-commerce, product design, UI/UX and gaming.
Seedance 2.5On-screen text and logos render accurately.
Seedance 2.5Use first and last frames for product showcases.
Seedance 2.5Stable dialogue in 11 languages.
Seedance 2.5Reference images keep characters consistent.
Seedance 2.5Transfer motion from one video to another (V2V).
What MiniMax H3 supports today: 4–15 second clips at 24 fps, 768p by default with an optional 2K regenerate step, 32 kHz stereo audio, aspect ratios from 21:9 to 9:16, and dialogue in 11 languages.
| Spec | MiniMax H3 |
|---|---|
| Duration | 4–15 s |
| Resolution | 768p default, up to 2K |
| Frame rate | 24 fps |
| Aspect ratios | 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 |
| Audio | 32 kHz native stereo |
| Inputs | text, ≤2 frames, ≤9 images, ≤3 videos, ≤3 audio |
| Dialogue languages | 11 stable |
| Weights | open (Hugging Face, 33B) |
Plans & Credit Packs
First-month BOGOStart your first paid subscription with any monthly plan and get 100% extra credits in month one.
For accounts that have never had a successful paid subscription. All monthly plans qualify; once per account. Bonus credits expire at the end of the first billing period.
First-month BOGO
300 credits×2
600 credits in your first month
Then 300 credits per month at the same monthly price.
For hobbyists and new users who want lightweight creation, personal testing, and a cost-effective way to keep producing steadily.
What you get
Actual output varies by model, resolution, duration, and reference media.
First-month BOGO
660 credits×2
1,320 credits in your first month
Then 660 credits per month at the same monthly price.
For most creators who want strong value, regular content updates, frequent generation, and marketing asset production.
What you get
Actual output varies by model, resolution, duration, and reference media.
First-month BOGO
1,450 credits×2
2,900 credits in your first month
Then 1,450 credits per month at the same monthly price.
For professionals, studios, and high-intensity creation needs. A production-ready plan for agencies and creative teams.
What you get
Actual output varies by model, resolution, duration, and reference media.
Answers about model selection, text-to-video, image-to-video and writing prompts.
Yes. New accounts get 6 free credits to try the AI video generator.
MiniMax H3 is a general-purpose multimodal video generation model from MiniMax. It understands text, images, video and audio and generates 4–15 second videos with native stereo sound at up to 2K.
MiniMax announced H3 on July 31, 2026 and open-sourced its weights on August 3, 2026.
Yes. MiniMax published two base checkpoints, H3-Base-FL2VA and H3-Base-Ref2VA, on Hugging Face (MiniMaxAI/MiniMax-H3). 2K output uses an additional H3-Regenerate-2K step.
[Free credits TBD] New accounts get free credits to generate a MiniMax H3 clip in the browser, with no credit card.
Write a prompt or upload first/last frames or references, choose duration (4–15 s) and aspect ratio, then generate and download the MP4 with audio.
Text only; one or two frames (first, last, or both); or omni reference with up to 9 images, 3 videos and 3 audio clips (audio must accompany an image or video).
Dialogue is stably supported in 11 languages: Arabic, Chinese, English, French, German, Italian, Japanese, Korean, Portuguese, Russian and Spanish.
MiniMax lists H3 in its platform model release notes; check platform.minimax.io for current API access and pricing.

Enter a prompt or upload a frame, add sound or dialogue, and generate a 4–15 second video with native stereo audio.
Create with MiniMax H3