Compose
H3 ref2va · audio + video0 / 6000
References none
Optional conditioning. Up to 9 pictures, 3 videos, 3 audio clips (16 MB each).
Refer to them in the prompt as <Picture 1>, <Video 1>, <Audio 1>.
Opening frame MOSS-Image → fl2va none
Generate a still, then animate it as shot 1. A keyframe replaces the picture/video/audio references above, and the prompt should describe the motion you want. checking MOSS…
No frame yet — the clip opens from the text prompt.
Dialogue Drama 3 none
A voice-over, not lip-sync. Generate the line now; H3 muxes it during the render. Owner-only: the hosted Drama 3 licence covers personal and evaluation use. checking Drama…
Voice reference (optional) none
No voice yet — the clip keeps its H3 soundtrack.
Clips upload straight to s3://bucket1-708/prism/studio. Long jobs keep running if you close the tab.
No clips yet.
Write a prompt and hit Generate — renders appear here with live progress.Nothing in this prefix.
Where it goes
Credit balance
remaining = allowance − credits appliedModal reports how many credits were applied this cycle, not the plan balance. Set the allowance once and the remaining figure is calculated from live billing data.
Story batch
one prompt per lineOpening frame shot 1 only none
Generate a MOSS still for the first shot. Shot 1 opens from it; later shots continue from the previous shot's last frame. Needs Continuous.
No frame yet — shot 1 starts from the text prompt.
Dialogue script one line per shot none
Optional. Use one line per shot; leave a line blank for a silent shot. Every voice is generated before the chain starts, so the GPU never waits on Drama 3. The voice reference from Compose is reused.
Uses the quality, steps and aspect from the form on the left. Without chaining, the attached references stay fixed for the whole batch. With Continuous, shot 1 opens the story and every later shot starts from the previous shot's last frame, all in one warm container (references are not used while chaining).
Image batch
MOSS-Image · one still per rowRenders stills on their own, outside the video queue. Every row can override every setting; a blank field inherits the shared defaults. Items run one at a time and keep rendering if you close the tab.
Advanced defaults
Skip layers default to 22. Leave SLG layers blank to disable it.
Speech batch
Drama 3 · one line per rowGenerates speech only — not lip-sync. Every row can override every setting; a blank field inherits the shared defaults. Lines run one at a time and keep rendering if you close the tab.
Reference audio optional · shared none
Upload one clip to clone its voice. Every row can opt out or use its own S3 key in the advanced row settings.
Prompt recipes
One click fills the composer with a prompt and the settings it was written for. Picture, multi-reference and audio recipes expect the matching reference in the tray; the card marks it red when it is missing. Full stories load every shot into Batch. Read the full guide →