AI Voice Cloner: Generate Speech from a Voice Sample

Turn a voice recording and new script into a WAV take you can review. Upload one audio sample, enter up to 4,000 characters in Text to Speak, and generate speech using the sample. A clear 5–10 second recording of one speaker is a practical starting point. This workflow uploads your recording for online processing and requires sign-in and an internet connection. Listen to the result and check pronunciation and pacing before using it.

The recording provides a voice sample, and Text to Speak provides the new words. Process them in the workspace to generate speech you can listen to and download as a WAV file. Treat the first result as a take to review: check its pronunciation and pacing against your script.

• Use AI Voice Cloner when you have a voice recording and want new words spoken using that sample. The recording supplies the voice sample; your script supplies the words.
• If you don’t have a reference recording, use Turn your script into spoken audio to choose a preset voice and generate speech as a WAV file.

1. 1

   Open AI Voice Cloner in its linked workspace. This online workflow requires sign-in and an internet connection.

2. 2

   Upload one recording in .mp3, .wav, .ogg, .flac, .m4a, or .aac format, up to 25 MB. Start with a clear 5–10 second sample from one speaker; use your own voice or a voice you have permission to use. In Text to Speak, enter up to 4,000 characters. Set Speed between 0.5× and 2.0× in 0.1 increments; its default is 1. Quality offers 16 (Fast), 32 (Balanced), and 64 (High Quality), with 32 as the default. Voice Accuracy ranges from 0.5 to 3 in 0.1 increments, with a default of 2. Review the visible controls and restrictions before processing. The listed extensions do not establish support for every codec or bit depth.

3. 3

   Click Process to generate speech from the recording and script. Play the result and check names, numbers, pronunciation, and pauses against your text. If a line sounds wrong, revise it and generate another take. Download the WAV when ready, then verify the downloaded file before using it.

More ways to use it

• A presenter with a reference recording.: They want to create a short spoken introduction using new words and a voice sample they have permission to use. A WAV take to listen to and check for pronunciation and pacing before use. Text to Speak: A short introduction within the 4,000-character limit., Speed: 1.
• A creator preparing a narrated video.: They want a speech track from a script and voice sample, then may need a separate tool to animate a portrait or add subtitles to a video. A WAV speech file to review. To continue toward a narrated video, create the voice track here, then use a separate step to animate a portrait and add subtitles. Add Audio to Video replaces the original soundtrack; it accepts one video and one audio file, up to 500 MB each, and produces an MP4. Text to Speak: A short narration passage within the 4,000-character limit., Speed: 1.

A name or number in the generated line sounds different from the script.

Try this: Correct the wording in Text to Speak, process another take, and listen for that name or number.

What to expect: The next take is generated from the revised script. Check how the name or number is spoken.

The generated speech pace doesn’t suit the passage.

Try this: Choose another Speed value within the available 0.5–2.0 range and process a take to review.

What to expect: The next take uses the selected Speed setting. Listen to decide whether the pacing suits the passage.

Online processing uploads files. Completed results may remain available for later download. See Privacy Policy for current file-handling and retention details. Check the processing mode before starting.