Turn a clip's captions into narration with an AI voice

Sometimes the words in a clip are right and the recording is wrong. The room echoed, the delivery was uneven, or you want a different voice entirely. DepCut can read a clip's captions aloud with an AI voice and lay the result over the clip. You fix the words once and get new audio from them.

Get the words down first

1. Open the Subtitles tab and press Generate subtitles. Transcription runs on a hosted speech service and spends credits.
2. Open Content to see the transcript. Each line is editable text with a timestamp.
3. Correct the wording right in the transcript. Press Enter inside a line to split it, and Backspace at the start of a line to merge it with the one before.

The AI voice reads exactly what the captions say. A misspelled name stays misspelled, so scan the transcript for names and numbers before you generate.

You can also skip the separate transcription step. If the clip has no captions yet, the next step transcribes it first.

Voice the clip

1. Click the clip on the timeline.
2. In the right panel, open the Volume tab.
3. Under the clip volume and Mute audio controls you find a voice menu titled Generated audio. Open it and hover a portrait to hear a voice, then click to pick it.
4. Press Generate audio for clip.

The button reads only the captions that overlap this clip, so the rest of your cut is untouched. If the clip has no captions, DepCut transcribes it, then voices the new lines. If it finds no speech, you get a short "No speech found to voice" message.

The voice uses DepCut Voices and your account credits. The same voice and language settings are shared with the Audio tab.

What lands on the timeline

The reading lands on the soundtrack, starting at the first caption, and it ducks the rest of the audio the way a new voiceover does. DepCut also re-times your captions to the new audio. The AI voice paces the lines differently from the original recording, and the re-timing keeps each caption, and its word highlight, in step with what is being said.

Once generated audio overlaps the clip, a Volume slider appears under the button. It controls every generated voice that overlaps the clip, so you can balance it against the clip's own sound from one place.

Replace the original sound

The original audio still plays under the new voice, at a lower level because the reading ducks it. To hear only the new reading, switch on Mute audio in the same Volume tab. The clip keeps its picture and loses its own sound.

Play the clip from the start. Check that the new reading fits the length of the picture and that no line runs into the next shot. If a line runs long, tighten the caption text and generate again.

When this works well

• A clip whose wording needs a fix: correct the transcript and regenerate.
• A clip recorded in a noisy or echoey room.
• A tutorial where one consistent narrator should cover several recordings.
• A talking segment you want delivered in a calmer or crisper voice.

For a voice you direct yourself, with delivery notes and a choice of models, use the Voice tab in the Audio tab. This button is the quick route when your words already live in captions.

Start a project

Open DepCut, generate subtitles for a clip, and press Generate audio for clip. Compare the new reading with your original and keep the one that sounds better.