Lip-sync dialogue video prompt
Write dialogue shots that models with speech generation can deliver cleanly: one speaker, short lines, clear delivery notes.
Fill in your details
Values are inserted into every prompt below. Nothing you type leaves your browser.
11 field(s) left — unfilled variables stay as {{placeholders}}.
1Fill and generate
If your model supports speech, put dialogue in quotes and attribute it to one visible speaker. Short lines (one or two sentences) sync best. Face the speaker towards the camera, mouth clearly visible.
{{shot}} of {{speaker}} in {{setting}}. The speaker looks {{eyeline}} and says, in a {{delivery}} voice: "{{line}}" Only this one person speaks. Lip movement matches the words. Camera: {{camera}}, face clearly visible and well lit. Lighting: {{lighting}}. Background sound: {{ambience}}, quiet under the voice. Duration {{duration}}, aspect ratio {{aspect_ratio}}. Avoid: text, captions, watermarks, logos, extra limbs or fingers, warped faces, flicker, sudden cuts, subtitles, other people talking, music over the dialogue, mouth moving after the line ends.
Example input
Elderly woman in a farmhouse kitchen, teasing line about storms, candlelight; 6s; 16:9.
Example output
Medium close-up of an elderly woman with silver braids in a green wool shawl in a candlelit farmhouse kitchen. The speaker looks just off camera to the left, as if at a grandchild and says, in a warm, slow, slightly teasing voice: "You think the storm scares me? I've seen forty winters worse than this." Only this one person speaks. Lip movement matches the words. Camera: static, very slight push-in, face clearly visible and well lit. Lighting: warm candlelight from the right, soft shadows. Background sound: wind outside, a crackling fire, quiet under the voice. Duration 6 seconds, aspect ratio 16:9. Avoid: text, captions, watermarks, logos, extra limbs or fingers, warped faces, flicker, sudden cuts, subtitles, other people talking, music over the dialogue, mouth moving after the line ends.
Troubleshooting
- The line is cut off: shorten it; aim for about 2.5 words per second of clip.
- Wrong person speaks or two voices appear: keep one person in frame for dialogue shots.
- The model adds burned-in subtitles: add 'no subtitles, no captions' at the start as well.
Known limitations
- Speech generation availability, languages and quality depend on the model and plan; check your tool.
- Never generate dialogue in the voice or likeness of a real person.