How to use the splitter
- Paste your script. Add visual directions in [square brackets] where you already know what should be on screen.
- Choose how to split: every N seconds suits fast-paced faceless videos; paragraphs suit slower documentaries.
- Set the voice pace so the timings match your narration.
- Pick an image style and aspect ratio, and optionally a setting to add to every prompt (for example a place and period).
- Copy the prompts into your image generator, or download the CSV as a shot list.
Getting better image prompts
- Describe what is seen, not what is said. "Rumours spread about the emperor" is narration; "[Crowd whispering in a Roman street at night]" is a picture. Bracketed cues give much better prompts.
- Keep one setting for the whole video. The optional setting field keeps period, place and mood consistent from scene to scene.
- Use one style per video. Mixing styles looks unplanned.
- Check history and people. Image models get dates, uniforms and faces wrong. Review every image before it goes in the video.
How long should each scene be?
There is no fixed rule. Shorter scenes feel faster and need more images; longer scenes suit calm narration and slow camera moves over each image. Work out the total with the script length calculator, and record the narration with the teleprompter or generate it with the voice generator.
FAQ
How does the splitter decide where a scene ends?
You choose: one scene per sentence, one per paragraph, or every N seconds of narration. In the seconds mode it groups whole sentences until it reaches the target length at your words-per-minute pace, and only cuts inside a sentence if that sentence alone is much longer than one scene.
Are the image prompts written by AI?
No. They are templates built from your text: the visual cue in [square brackets] if the scene has one, otherwise the narration, plus the style and aspect ratio you choose. Nothing is sent to an AI model or a server. Edit them before you use them; a good prompt describes what should be seen, not what is said.
How are the scene timings calculated?
Each scene lasts its word count divided by your words-per-minute pace. Text in [square brackets] takes no time. Real voiceovers vary, so check the timings against your audio.
What can I do with the CSV?
Open it in Google Sheets or Excel as a shot list, or import it into any tool that reads CSV. It has the scene number, start and end time, length, word count, narration, visual cue and image prompt.
How many scenes does a long video need?
Divide the length by the seconds per scene. A 30-minute video at 8 seconds per scene is 225 scenes; at 5 seconds it is 360. That is why long-form faceless videos usually need generated or batch-produced visuals.
Is my script stored or uploaded?
No. Everything runs in your browser and nothing is saved or sent anywhere.