3 · Visuals

The Visuals screen turns a finished voiceover into pictures. It has two views: Generate visuals, where you configure the run, and Generated scenes, where you review what it planned and generate it.

Settings

Source and models

Source picks the Project and which Voiceover to use. Models sets the Default image model, with its per-image credit cost shown, and the Prompt model that writes the prompts.

Style & content

Visual guidance is a free-text field describing how images look and what they show — there is no fixed art-style menu. Browse styles opens the Style library, a dozen proven looks for faceless channels; Use this style replaces your guidance and image model with that style's. Generate from images builds a style from reference images you upload instead.

Pacing

Simple takes a Number of scenes (optional) and a Minimum scene duration (seconds), and makes every scene an AI image. Advanced opens a timeline you cut into segments, each with its own scene count and its own mix of AI image, AI video and Stock footage, plus a toggle that lets the app draw custom motion graphics for standout moments. A segment set to zero scenes holds the previous visual through that stretch.

Characters

Turn on Characters to keep a recurring cast consistent across scenes and videos. Generate Characters reads the script for recurring characters; Upload Custom Character adds one from an image. Each has an editable appearance, a Role in the video, and a presence of AI decides or Always. Your plan caps the roster size and how many characters one scene can hold.

Prompts and scenes

The footer button — Generate image prompts, with the scene count in it — splits the voiceover into scenes and writes one prompt per scene, then moves you to Generated scenes.

Each card shows the scene number, its timestamps, who is cast in it, the narration it covers and its Image prompt. Copy the prompt, edit it inline, or use the sparkle to regenerate it with a What should change? instruction; the scene keeps its current image until you regenerate that separately. The cast badges open Characters in this scene, which rewrites the prompt to match who you pick.

Generating and moving on

Generate two or three scenes first to check the look. Generate all visuals then shows how many images, clips and stock shots the run needs and the Estimated charge, and warns that generation cannot be cancelled once it starts. Individual scenes keep their own Generate image or Regenerate button with a model menu beside it, and anything that failed gets Retry or Retry all failed.

Generation status reports progress while it works, and Continue to render appears when the run is done. Generated visuals are stored with the project for the number of days the page states.

Save the whole configuration — look and pacing together — as a preset from Presets in the Models header, and tick Set as default for this channel. Next: Render.