2 · Voiceover

The Voiceover screen reads a project's script aloud. The narration is the clock for the rest of the pipeline: scene boundaries, visuals and subtitles are all timed to it, so this step comes before visuals.

Already have a recording? Skip this screen and upload it instead: Use your own script or voiceover.

Pick the project

The Project panel has a Choose a project dropdown, with your newest projects as chips underneath and a Show script button to check what will be read. If you have text that is not a project yet, Create voiceover from text takes a paste and makes one.

Choose a voice

Voice settings opens on Standard and can switch to Voice Clone.

• Standard — pick a Provider (ElevenLabs or Chatterbox, which is English only), then a Voice. Select voice opens the library: search by name or description, filter by All, Female or Male, and play a sample from any row. ElevenLabs also exposes a Model choice, and a Manual voice ID (optional override) if you want a voice that is not in the list.
• Voice Clone — pick a Clone provider and one of your clones. Add voice clone opens Add Voice Clone: give it a Name and upload one clear Reference Audio sample, within the length and file-size limits the dialog states. You can hold a small number of clones at once, rename or delete them, and test one on a short line before committing to a full run.

Which to use

For most channels we recommend a voice clone: it is far cheaper per minute than a standard voice, and it is what many large faceless channels run on. The trade-off is that cloning takes some experimenting. The model suits some voices better than others, and the sample matters as much as the voice. If a clone sounds off, try a different reference recording first, then a different voice altogether, and use the short test line before spending on a full run. Standard voices cost more but give a predictable result on the first attempt, which makes them the safer pick for a one-off video.

Speed is a slider around a default of 1.0x. On ElevenLabs, Advanced voice settings adds Stability, Similarity, Style exaggeration and Enable speaker boost, with Reset to defaults to undo the lot. With an ElevenLabs key stored in Account & Settings, Use own key bills your ElevenLabs account directly instead of spending credits; see Credits and usage.

Options before you run

• Speak chapter titles — reads later section titles aloud. Only appears when the script has more than one.
• Loudness normalization — evens out volume across a long narration. It adds a stated percentage to the cost, and the app suggests it afterwards if a voiceover came out quiet.
• Limit to first N words — generates only the opening words, for testing settings cheaply.
• Open visuals generation page once finished — jumps straight to the next step.

Generate voiceover opens a confirmation showing provider, voice, speed, word count and the Estimated cost in credits. A voiceover task cannot be cancelled once started, and a failed one is refunded.

What you get

Generation runs in the background — leave the page, or follow it from View voiceover jobs. When it lands you get Voiceover ready with a player, Download Audio and Continue to Visuals. The audio, its subtitles and word timings are saved to the project and kept for the number of days the card states, so download a copy if you need it long-term.

Save the settings as a Voiceover preset and set it as the channel default to skip this configuration next time. Next: Visuals.