You describe a video shot by shot. We generate each clip, keep your characters consistent across all of them, add voice and music, and stitch the result into one finished file — up to five minutes long.
A character is a person you want to look identical in every shot. Upload 4+ photos, then train them once.
Training costs 20 credits and takes about 12 minutes. Use clear photos where the face is plainly visible — blurry or half-empty frames make the result worse. This is the thing generic clip generators can't do. Optional: skip it and you still get video, you just won't get the same face twice.
A scene is the place and film style your shots share. Write it once, apply it to many shots, and they feel like one film instead of unrelated clips.
✓ “Sunlit school field, warm 35mm film, shallow depth of field, dust in the air” — that's a look. ✗ “Maula cautions Akula” — that's something that happens, so it belongs in a shot inside your project, not here. Rule of thumb: if it could describe twenty different moments, it's a scene; if it happens once, it's a shot. Also optional.
One project is one video. Paste your script and we cut it into 10-second shots, one per sentence, attaching any character whose name appears in the line.
You can also add shots by hand. Write like a director: “close-up, golden hour backlight, she turns and smiles” beats “a woman smiling”. If you skip the detail we add sensible camera and lighting language for you. Nothing is generated — and nothing is charged — until you press Generate.
Use Draft to block out composition cheaply, then re-render the keepers at Standard.
Draft is 1 credit/second and takes about a minute. Standard is 8 credits/second at 720p. Maximum is for hero shots only — it looks slightly better but takes around 30 minutes.
Type a line under any shot and pick a voice. Choose a music track — we suggest one based on your shots.
Voiceover takes under a minute per line. Music plays under the whole video and is automatically ducked beneath the voice so words stay clear.
Hit Render video. Everything is stitched into one MP4 with audio, ready to download.
Rendering takes seconds once your clips exist. Watch progress in the sidebar — you can keep working while it runs.
| Action | Credits | Time |
|---|---|---|
| Draft clip (480p) | 1 / sec | ~1 min per 5s |
| Standard clip (720p) | 8 / sec | ~2 min per 5s |
| Maximum clip (720p) | 20 / sec | ~30 min per 5s |
| Train a character | 20 | ~12 min |
| Voiceover line | free | under a minute |
| Music & final render | free | seconds |
About 36 shots. On Standard that's roughly two minutes of generation per shot, and shots queue in parallel. Budget an afternoon for a polished 3-minute piece — most of which is you reviewing and re-rolling shots you don't like.
Trained characters apply on Draft and Standard. Maximum uses a larger checkpoint the training doesn't transfer to, so character shots should use Standard; save Maximum for scenery.
Face clearly visible, well lit, varied angles and expressions. Four is the minimum; eight to twelve is better. Photos where the person is small, blurry or turned away actively hurt the result.
Yes. Every track in the library was generated on our own hardware from our own prompts, so there's no attribution requirement and no third-party licence attached to your video.
Hit Regenerate. Failures are usually transient. You're only charged for clips that succeed.
Ready to make something?
Start with a character if you want them to recur, or jump straight into a project.