MiniMax H3 Optimised Text to Video

MiniMax H3 Optimised Text to Video

What is MiniMax H3 Optimised Text to Video?

MiniMax H3 Optimised Text to Video turns a single written prompt into a short clip that already has sound. It serves MiniMax's H3 (Hailuo 3.0) model — the omni-modal generation model that reads your prompt as one context and returns picture and audio together — on an optimised inference stack for faster turnaround at matched quality. Describe a scene, a subject, a camera move, a line of dialogue and the sounds you want, and the model renders a 5-to-15-second video at 24fps with synchronized stereo audio in the same pass, with no separate voice-over or sound-design step.

Key Features

• •Native synchronized audio: dialogue, sound effects and ambience generated with the picture and timed to the on-screen action, with lip movement matching the speech.
• •Text-to-video from a plain prompt, following director-style instructions literally.
• •Clip lengths of any whole number of seconds from 5 to 15, at a film-standard 24fps.
• •Resolutions from 480p to 2k — 768p is generated natively, while 1080p and 2k are upscaled.
• •Aspect ratios for every destination: 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9.
• •A quality switch (high or low) and an optional prompt-enhance pass.

Best Use Cases

Short-form and social clips, dialogue-driven scenes, trailer and TVC concepts, product and brand films, and storyboard or previsualization shots where you want motion and sound without a separate audio pass. Because audio is native, the model is strong for talking-character moments and for scenes where a specific sound effect has to land on a specific beat. The 24fps output cuts straight into an editing timeline.

Prompt Tips and Output Quality

Treat the prompt like a shot brief: name the camera move, then the subject, then the action, then the scene, lighting and mood as one flowing sentence, and spell out both the spoken line and the sounds you want. Keep dialogue to one short, clear line in shorter clips — lip-sync can drift when many lines are packed into a few seconds. In testing, a 768p five-second clip of a steam-locomotive driver calling a line as the whistle shrieks came back clean, with the spoken line lip-synced and the whistle and steam landing on the action. Avoid asking for legible on-screen text, which is the weaker area on this stack.

Using MiniMax H3 Optimised Text to Video

Write the prompt like a shot brief

H3 follows instructions literally, so give it a blueprint rather than a keyword list. A reliable order is camera shot and motion, then subject and description, then the action, then the scene, lighting and mood — expressed as one flowing sentence. Because audio is generated with the picture, write the sound in too: the exact spoken line in quotation marks, and the effects or ambience you want to hear.

Get clean dialogue and sound

Keep spoken lines short and clear, especially in 5-to-7-second clips, because lip-sync can drift when several lines are crammed into a few seconds. Put one clear line in each clip and let a sound effect land after it rather than over it. For longer 12-to-15-second takes, add continuity cues so the shot holds together from the first frame to the last.

Choose resolution and length

768p is the only natively generated tier and makes the best default; 1080p and 2k are upscaled, so reach for them only when you need more pixels. Pick the shortest duration that covers your beat — 5 seconds for a single action, up to 15 for a short multi-shot sequence. Match the aspect ratio to where the video will run: 16:9 for widescreen, 9:16 for vertical social, 21:9 for cinematic.

Recommended Settings

• •Dialogue-driven clip (tested): resolution 768p, duration 5, aspect_ratio 16:9, quality high, enhance_prompt on. A five-second steam-locomotive scene with a spoken line and a whistle came back clean at 1344x768, 24fps, with stereo audio.
• •Fast iteration: set quality to low to preview motion and timing, then switch to high for the final take.
• •Finished prompt: turn enhance_prompt off to send your prompt exactly as written.
• •Vertical social: aspect_ratio 9:16 with a duration of 5 to 8 seconds.

Examples to try

Each example ran end to end from the prompt alone at the settings shown, carrying the spoken line, the sound and the camera move into the clip.

• •Native dialogue and synced sound in one pass. Prompt: "On the footplate of a steam locomotive pounding along an open track, a soot-streaked driver in a striped cap and grey overalls leans out of the cab window, looks back along the train and says clearly, 'Right, she's holding pressure - take us up the bank.' The firebox throws an orange glow across his face, white steam streams past, and a moment later the whistle lets out a long shriek as the wheels hammer the rails. Medium shot from alongside the moving cab." The line is lip-synced and the whistle lands on cue at 768p/5s; keep the spoken line to one short sentence so the sync holds.
• •Native dialogue in a second language. Prompt: "Medium close-up from across the counter, the camera holding steady. A sushi chef in a white jacket and black bandana presses a piece of salmon nigiri between his palms, sets it down on a dark lacquer board, then looks up to the camera and says clearly in Japanese, 'どうぞ、召し上がれ。' Behind him a cedar counter and a row of glass fridges; the knife clinks on the board and low restaurant chatter fills the room." The chef's mouth forms the Japanese words over restaurant ambience; put the exact foreign line in quotes and keep it short.
• •Holds a longer continuous take. Prompt: "A steady push-in from a wide frame to a close shot of the hands. At a wooden handloom strung with indigo and cream threads, a weaver in a linen apron throws the wooden shuttle through the warp, pulls the reed toward her to beat the weft tight, then throws the shuttle back the other way as a striped band of cloth grows row by row on the loom. Daylight from a side window catches the lint in the air; the loom clacks and the treadles thump in a steady rhythm." One unbroken 10-second push-in holds from the wide loom to the hands; name the camera move and add a continuity cue for longer takes.

Built for more: try "a street violinist says 'last one tonight' as coins drop in the open case", or "a 12-second walk-and-talk through a misty greenhouse, leaves brushing the lens."

Does it really generate sound?

Yes — dialogue, sound effects and ambience are produced in the same pass as the video and timed to what happens on screen.

How long can clips be?

Any whole number of seconds from 5 to 15, at 24fps.

What resolution should I use?

768p is generated natively and is the recommended default; 1080p and 2k are upscaled. 4k is available on the API only and returns HEVC, which many browsers play as audio only.

Is MiniMax H3 the same as MiniMax M3?

No — M3 is MiniMax's language model, while H3 is the video model.

Can I make vertical video?

Yes — choose the 9:16 aspect ratio for vertical social formats, or 21:9 for cinematic framing.