Cinematic video with consistent characters and real voices
MiniMax H3 turns a prompt plus up to nine reference images and three voice samples into cinematic clips of up to 15 seconds — keeping faces, outfits and places consistent from shot to shot. It powers OkayPic's short-drama workflow.
from $0.007/s pay as you go
Give it characters, places and props as pictures, and voices as audio. H3 keeps who is who across every shot of the clip.
Pin the opening frame to an image you approve, then let references drive the rest — composition stays exactly where you put it.
Set where the clip starts and ends and H3 fills in the motion between them.
Drive a talking shot from your own audio, up to 15 seconds.
API
Call MiniMax H3 with model id minimax-h3. Same price as the web app.
POST https://api.okaypic.com/api/video/generate
{
"model": "minimax-h3",
"prompt": "…",
"duration": 5
}From $0.007/s at 480p — a 10-second clip costs under ten cents. 768p and 1080p are priced per second too; see Pricing.
1 to 15 seconds (up to 10 seconds at 1080p).
No. OkayPic is pay as you go: top up any amount and pay only for what you generate. Failed generations are refunded automatically.
Yes. Every model on OkayPic is available through one REST API with the same prices as the web app. Create a key on the Developer page and see the API docs.
Yes, what you generate is yours to use, subject to the model provider's content policy.
From $0.007/s. No subscription — failed generations are refunded.