High-quality production · Pro only
Narrated Story · Shorts 9:16 · Longform 16:9
Carry a story from the first scene to the last without the seams showing. You walk
nine steps yourself, but navigation and generation are separate, so moving between
steps never regenerates what you already made and you can click the step indicator
to jump straight to any step. Shorts run 4, 6, 8, or 10 scenes; longform runs 6,
10, 15, 20, or 30.
← Back to the main page
- Script
- Voice
- Planning
- Characters
- Images
- Video
- Subtitles · BGM
- Review
- Render
Highlighted steps are the ones that decide the result in this mode.
Shot, angle, move, and lighting from dropdowns
In the planning step every scene gets a shot size, camera angle, camera move, and lighting from a fixed vocabulary. AI fills in a draft; to fix a scene you change a dropdown instead of rewriting a paragraph. The app warns when the same shot repeats too many times in a row.
Lock the cast with character sheets
A persona is one presenter; character sheets are the people inside the story. Several can appear in one video, the front-face panel anchors the same face and outfit in every scene, and sheets are saved to a library for the next project.
Each scene picks up where the last one ended
Scenes are planned ten at a time, and the ending state of the last scene in one batch is fed into the next, so story and picture do not jump at batch boundaries. The video prompt is designed separately: how this still image moves until it reaches the next scene's image.
4-15 second clips with MiniMax H3
Clip length follows the narration, and the previous scene's last frame is passed in as the starting reference for the next clip. Narration that overruns 15 seconds is rewritten shorter for that scene only. H3 adds sound effects, while background music is left to the BGM you choose in the app.
Only the missing pieces get regenerated
If one of eight images failed, only that one is redrawn. For images, video, and voice alike, the app asks what to generate when some scenes already exist, and a "needs attention only" filter narrows a 30-scene longform down to the scenes that need work.
Autosave and the High-Quality Library
Edits are saved as you type, not just when you change steps, and the last save time stays on screen. Failures are reported as one plain-language next action, such as "your credits are exhausted", and the High-Quality Library reopens any project.
Sample · Narrated Story longform 16:9 · straight out of this mode
Video not playing? Watch on YouTube
⚠️
Costs are announced up front
High-quality production mode needs both an OpenAI key and a MiniMax key. The
H3 video model is billed per second, so cost grows with duration and scene
count; the video step header shows the seconds about to be billed, and you can
generate only the scenes you select. See the
API cost guide.