hermoso
Login Sign up

Models › Video models

MiniMax

MiniMax H3 References

Up to 9 reference images (a place, a character, a prop) in one clip, up to 2K

MiniMax · video model

5–15sclip length
4kbest resolution of 4
Native soundspeech and world audio in the render
up to 9reference images

MiniMax H3 References is one of 35 video models Hermoso can render a video ad, product sizzle or UGC clip with. Name it, or leave the model out and the router picks a sound default. The same catalog is available from the app, the MCP server and the CLI, on one account and one credit balance; the exact price of every render is shown in the app before it runs.

What people make with it

The linked formats are creative recipes Hermoso plans and renders for your brand.

About MiniMax

MiniMax makes the Hailuo video models and the Speech voice line.

Specs

Lengths5s8s10s12s15s
Longest single take15s in one unbroken shot
Resolutions1080p720p480p4k
Aspect ratios16:99:161:14:33:421:9
SoundNative audio
Modestext to video, image to video, reference to video
Reference imagesup to 9 per render in Hermoso (our fidelity cap, measured on paid renders; the model itself accepts more), addressed in the prompt as @Image1, @Image2…
Edit an existing videoNo
Model idminimax-h3-ref

Where it fits

Each aspect ratio MiniMax H3 References renders maps to a placement:

  • 16:9 for YouTube, in-stream and display
  • 9:16 for Reels, TikTok, Shorts and Stories
  • 1:1 for Feed posts and carousels
  • 4:3 for Presentations and display
  • 3:4 for Pinterest and tablet feed
  • 21:9 for Cinematic banners and hero films

Just ask for it

In Claude, ChatGPT, Cursor or any assistant connected to Hermoso, name the model in plain English. Or pick it from the model menu in the app.

Make a 15 second UGC ad for our olive oil with MiniMax H3 References, kitchen at golden hour, 9:16.
Render the packshot spin on MiniMax H3 References and post it to TikTok and Reels.
Redo yesterday's ad on MiniMax H3 References with the new label.

Getting the best out of MiniMax H3 References

  • Describe one continuous camera move and where it ends. The model renders motion, not a shot list.
  • Attach the real product photo as a reference and ask for it matched exactly. Never describe a lookalike.
  • Name colours in words. A hex code turns into text on screen.
  • Keep spoken lines short. Roughly two words per second is what a clip can carry cleanly.
  • Ask for the placement's ratio up front (9:16 for Reels and TikTok) instead of cropping afterwards.

Make a video ad with MiniMax H3 References. Start free, and every feature is on every plan.

Start free →   See pricing

Frequently asked questions

How long a video can MiniMax H3 References make?

MiniMax H3 References renders clips from 5 to 15 seconds. Longer spots are built from several clips.

Does MiniMax H3 References make sound?

Yes. MiniMax H3 References renders native audio with the picture, so a talking creator or a voiceover is part of the same take.

Can MiniMax H3 References use my product photos?

Yes. Attach up to 9 reference images of the product and Hermoso keeps the real product in the render.

How much does MiniMax H3 References cost?

The exact credit price is shown in the app before every render and settled at cost. See pricing for what a credit is worth.

Related video models

Other MiniMax models first, then the video models listed nearest to MiniMax H3 References in the catalog.

MiniMax H3 Max References

Up to 9 reference images in one clip, such as a person and several products, all in the same scene; native audio, 480P to 1080P, up to 15s

5–15s · 720p / 480p / 1080p · native sound

MiniMax H3 Max Multi Angle

Moves the camera around one still image along a path you can set: orbit, rise or push in; 480P to 1080P, up to 15s

5–15s · 720p / 480p / 1080p · native sound

MiniMax H3 Max Turbo

The faster H3 Max at half the price; up to 768p and 15s; can go from a first to a last frame

5–15s · 720p / 480p / 1080p · native sound

Seedance 2.5 Edit

Region-level edits on a clip you already have: change one object, face or setting and the rest stays untouched · keeps synced audio · up to 30s

4–30s · 1080p / 720p / 480p · native sound

FLUX 3 Edit

Edits a clip you already have from a text instruction, keeping its motion, timing, framing and sound; clips under 15s

5–15s · silent

Kling O3 Edit

Edits a clip you already have from a text instruction and up to four reference images, keeping its motion, timing and sound; clips of 3 to 15s at 720px or wider

5–15s · silent

Seedance 2.5 default

Up to 30s in ONE unbroken take · synced audio · cannot take a photo of a person as a reference

4–30s · 1080p / 720p / 480p · native sound

MiniMax H3 Max

Tuned H3: follows prompts more closely and looks better; up to 768p and 15s; can go from a first to a last frame

5–15s · 720p / 480p / 1080p · native sound

See all 35 video models or every model Hermoso offers.