38 words. Spoken in one pass so the first word of each sentence stays intact.
1080 square. Captions and level bars are burned into the film.
Take 1 · 3.12 words per second · match 1.00.
| Take | Words/sec | Match score | Decision |
|---|---|---|---|
| 1 | 3.12 | 1.00 | Kept |
Makes a square social video narrated in a cloned house voice. Each recording is checked against the script, and captions and sound bars are animated in time with the words.
Recording a fresh voiceover for every post is slow. And a caption that drifts from the spoken line looks careless when the film is watched on mute.
A voice cloning model speaks the whole script in one go. A speech recognition tool listens back, and recordings that are too slow or do not match the script are set aside. The best recording is trimmed, then set with typed captions, sound bars and a music bed under the voice. A take report shows the pace and match score of each recording.
Interactive preview with sample data. Nothing is sent to a server.
38 words. Spoken in one pass so the first word of each sentence stays intact.
1080 square. Captions and level bars are burned into the film.
Take 1 · 3.12 words per second · match 1.00.
| Take | Words/sec | Match score | Decision |
|---|---|---|---|
| 1 | 3.12 | 1.00 | Kept |
flowchart LR
Script[Script] --> Clone[Qwen3-TTS]
Clone --> Read[Whisper]
Read --> Score{Pace and match?}
Score -->|Keep| Tight[Tighten pauses]
Score -->|Reject| Clone
Tight --> Mix[Music bed under the voice]
Mix --> Caps[Typed captions and level bars]
Caps --> Film[Square MP4 and poster]