Dialogue in 11 languages, including English, Chinese, Japanese and Korean
A tourist orders in English, she answers in English, and both mouths match the words
Start from a sentence, two frames or a few reference images and get 4–15s of 2K video with dialogue and sound effects built in
Free credits on signup · No model download or GPU needed
Three ways in, and every clip below was generated by MiniMax H3
No uploads needed: describe the subject, the action and the sound
From closed to open: the shutter rolls up and the lanterns light one by one
Up to 9 images, 3 videos and 3 audio files, 12 in total
Turn the sound on: the lines and lip movement come straight from the model
A tourist orders in English, she answers in English, and both mouths match the words
The prompt says MEI'S NOODLES, and the letters hold up as the sign goes up and the camera pushes in
Upload a clip, ask for a snowy night, and the people, motion and camera stay the same
Official 2K samples from MiniMax's H3 launch post; unmute to hear the native stereo
Scales, eye and skin texture stay sharp in a 2K close-up
Based on the options currently available on Pilio
Text, first frame, first and last frame, multimodal
Generate without uploads, or lock the opening and ending with images
4–15 seconds
Any whole second in that range
768p or 2K
768p by default, both at 24 FPS
Adaptive or 6 fixed ratios
21:9, 16:9, 4:3, 1:1, 3:4 and 9:16; first and last frame mode follows your images
9 images + 3 videos + 3 audio
12 files at most per run, and audio needs an image or video alongside it
Native stereo
Voice, effects and music are generated together, with stable dialogue in 11 languages
Free credits on signup, with the cost shown before you generate