MiniMax H3 · Native 2K Video With Synced Audio
Generate 5-15 second videos in native 2K with synchronized audio, natural dialogue and multi-shot storytelling — from text, images or reference material.
AI video examples on OVOV
Real generations from our video models — Veo 3.1, Kling and Seedance. Hover any clip to play, click the speaker icon to hear the audio.
Hover to play · Click speaker to unmute
AI video examples on OVOV
Real generations from our video models — Veo 3.1, Kling and Seedance. Hover any clip to play, click the speaker icon to hear the audio.
Alien Fantasy World
Luxury Perfume Ad
Product Launch Teaser
Maldives Travel Commercial
Eastern Wuxia Duel
Hollywood Racing Movie
Spaceship to the Stars
Golden Retriever & Butterflies
Sushi Chef at Work
Red Dress on the Cliffs
Midnight Duel
Handheld Selfie Vlog
Mouth-watering Food Close-up
Urban Street Dance
Luxury Lipstick Ad
Why MiniMax H3 stands out
A quality-first flagship: native 2K resolution, audio and dialogue generated in sync with the picture, and true multi-shot narrative in a single generation.
Native 2K Output
H3 renders at 2K natively — not upscaled from a lower resolution. Fine textures, readable text and crisp edges that survive big screens and post-production.
Synced Audio & Dialogue
Characters actually speak. H3 generates natural dialogue, ambience and sound effects synchronized with lip movement and on-screen action — one pass, no separate audio step.
Multi-Shot Storytelling
Describe a sequence of shots and H3 cuts between them with consistent characters, lighting and pacing — a complete narrative beat in a single clip.
Omni Reference Control
In text-to-video mode, feed up to 9 reference images, 3 reference videos (15 seconds combined) and 3 audio tracks to lock style, character and sound direction.
First & Last Frame Control
In image-to-video mode, set the first frame — and optionally the last — and H3 animates the motion between them. Ideal for reveals, transitions and brand-exact shots.
5-15s Flexible Duration
Choose 5, 10 or 15 seconds and any of 6 aspect ratios from 21:9 widescreen to 9:16 vertical. Credits scale linearly with duration, so costs stay predictable.
From idea to 2K video in three steps
H3 is quality-first — it takes a little longer than speed-focused models, and the output shows it.
Write your scene — with dialogue
Describe the shots, put spoken lines in quotes, and optionally attach reference images, videos or audio. Or start from a first frame in image-to-video mode.
H3 renders picture and sound together
The model generates visuals, dialogue and sound effects in one synchronized pass. Quality-first rendering may queue at peak times — if a generation fails, credits are refunded automatically.
Download native 2K, ready to publish
Export your clip in native 2K (or 768P for drafts) with no watermark and full commercial usage rights.
What creators build with H3
H3 shines wherever picture quality and believable sound matter more than raw speed.
Dialogue-Driven Ads
Spokesperson spots, testimonials and story ads where a character speaks to camera — lips, voice and gesture all in sync, no dubbing session required.
- Natural spoken lines from your script
- Native 2K holds up on TV and web
- Sound effects and ambience baked in
Short Films & Narrative Scenes
Multi-shot scenes with consistent characters and continuous story logic — establish, react, resolve — in a single 15-second generation.
- Shot-to-shot character consistency
- 21:9 or 16:9 cinematic framing
- Dialogue drives the drama
Social & Vertical Video
9:16 clips for TikTok, Reels and Shorts where crisp detail and real audio stop the scroll — talking heads, mini-skits, hook-first storytelling.
- 9:16, 3:4 and 1:1 vertical-friendly ratios
- 5-second hooks or 15-second stories
- Speech and sound ready for sound-on feeds
Product & Brand Showcases
Start from your exact product photo with first-frame control, end on the hero shot with last-frame control, and let H3 animate the reveal in 2K.
- First/last frame keeps the product on-model
- Up to 9 reference images lock brand style
- 2K detail flatters materials and textures
Prompt writing guide for MiniMax H3
H3 rewards prompts that read like a mini screenplay: shots, actions, and spoken lines in quotes.
Write the scene, quote the lines — H3 handles picture and sound
Put spoken lines in quotes
H3 turns quoted text into synchronized speech. Name who speaks, quote what they say, and describe the delivery.
- •Attribute each line to a character
- •Keep lines short — a 5-second clip fits one or two sentences
- •Add tone cues: whispers, shouts, hesitates
Example
"An old fisherman looks at the horizon and says softly: "The sea remembers everything." Waves lap against the hull, gulls cry in the distance."
Structure multi-shot sequences
Describe shots in order — establishing, action, reaction — and H3 cuts between them while keeping characters and lighting consistent.
- •One sentence per shot, in narrative order
- •2-3 shots fit comfortably in 10-15 seconds
- •Repeat character descriptors to anchor consistency
Example
"Shot 1: wide view of a neon street market at night. Shot 2: a chef flips noodles in a wok, flames leap. Shot 3: close-up of a customer's delighted face, steam rising."
Pick the right mode for the job
Text-to-video accepts multimodal references — up to 9 images, 3 videos (15s combined) and 3 audio tracks. Image-to-video starts from your exact first frame, with optional last frame.
- •Text-to-video + references: style and character control
- •Image-to-video: pixel-exact starting point
- •Add a last frame when the ending matters
Example
"Image-to-video: first frame is your product photo. Prompt: "slow 180-degree orbit, studio light sweeping across the surface, gentle whoosh sound"."
Match resolution and length to your budget
768P costs 25 credits per 5 seconds, native 2K costs 40 — and 10s/15s scale linearly at 2x/3x. Draft cheap, finish sharp.
- •Iterate prompts at 768P, render the keeper in 2K
- •5-second tests cost the least
- •Failed generations refund credits automatically
Example
"Workflow: three 768P 5s drafts (75 credits) to nail the prompt, then one 2K 15s final (120 credits)."
State the aspect ratio upfront
H3 supports 21:9, 16:9, 4:3, 1:1, 3:4 and 9:16. Composition changes with the frame, so declare it before describing the scene.
- •21:9 or 16:9 for cinematic and YouTube
- •9:16 for TikTok, Reels and Shorts
- •1:1 and 3:4 for feed posts
Example
"A 9:16 vertical clip: a barista looks up at the camera and says "Your usual?" — morning light through the cafe window."
Stack references deliberately
Each reference type does a different job: images carry style and identity, videos carry motion and pacing, audio carries mood. Use only what the shot needs.
- •Character images: 2-3 angles beat 9 random shots
- •Reference videos: 15 seconds combined, so trim to the moves you want
- •Audio references steer music and tone, not exact playback
Example
"3 photos of the same mascot + 1 dance clip + 1 upbeat track: "the mascot performs this dance on a rooftop at sunset"."