Text to video
No footage, no camera, no set. Describe the shot the way you would describe it to a cinematographer — subject, action, camera, light — and the model shoots it. Seedance 2.0 adds a synchronized audio track; MiniMax H3 renders at 2K.
Made with this model
Every clip below was rendered here, on the settings shown under it. The caption is the exact prompt that produced it.
Seedance 2.0 · 720p · 16:9 · 4s
MiniMax H3 · 2K · 16:9 · 5s
How text to video works here
Write the shot
Two or three sentences covering what is in frame, what moves, and how the camera behaves. Skip the keyword soup — these models read prose.
Frame it
Pick an aspect ratio for wherever the clip is going: 9:16 for vertical feeds, 16:9 for embeds and YouTube, 21:9 when you want it to feel like film.
Render
Four to fifteen seconds per clip. Progress streams live, and the finished MP4 downloads straight from the panel.
Specs at a glance
| Model | Duration | Resolution | Native audio | Reference images | From |
|---|---|---|---|---|---|
| Seedance 2.0 | 4–15 seconds | 480p · 720p · 1080p · 4K | Yes | Not used | 38 credits for 5 seconds |
| MiniMax H3 | 5–15 seconds | 2K | No | Not used | 130 credits for 5 seconds |
What separates a good prompt from a lucky one
Lead with the subject, not the style. “A ceramicist trimming a bowl” anchors the model; “cinematic, 8k, masterpiece” does not.
Describe one camera move and commit to it. Competing moves in one prompt produce a clip that cannot decide.
Say what time of day it is. It sets colour, contrast and shadow direction in a single word.
If you want sound, say what you hear. Seedance 2.0 scores the clip from the same prompt — ask for rain on a tin roof and you get rain on a tin roof.