Sarvam CreativeVoice Cloning (Beta)

Creating a Clone

View as Markdown

Clone a voice in three steps - record, personalize, and save - right in the browser. For an overview of Voice Cloning, its language support, and consent requirements, see the Voice Cloning overview.

You must have the rights to clone any voice you record. The final step requires you to explicitly acknowledge and consent to recording, processing, cloning, and using the voice - this isn’t a formality, it’s a mandatory checkbox before you can save the clone.

The Three-Step Flow

Voice cloning landing screen. Header: Voice cloning title, Feedback and History buttons. Center content reads Clone any voice in 3 steps with subtitle Powered by Sarvam's voice synthesis in 12 Indian languages. Below, three numbered steps with icons: 01 Record audio (microphone icon), 10 seconds of clear speech in the language you want to clone; 02 Processing (clock icon), Sarvam builds a voice model from your recording; 03 Use anywhere (waveform icon), Deploy your clone to use in Text to Speech. A black plus Clone a voice button sits below.

The three-step promise: record, process, use anywhere - set up front so you know exactly what to expect.

Record Your Voice

Clone voice modal, Step 1 of 3, titled Record your voice. A Language dropdown sits at the top. Below it, a Read this aloud box shows a short passage to read, its estimated read time, and a Shuffle button to swap in a different passage. A black Start recording button sits below the box, with the helper text Speak clearly, quiet room, no music beneath it.

Step 1: pick a language, read the provided passage aloud, and record - no upload required.
1

Start a New Clone

Click + Clone a voice from the Voice Cloning landing screen, the Voice Library, Dubbing’s speaker-strategy step, or Text to Speech’s Cloned voices tab.

2

Choose the Language

Select the language you’ll be speaking from the Language dropdown.

  • This is the language the clone is captured in - pick the one you’ll actually use it for.
3

Read the Passage Aloud

Read the text in the Read this aloud box, or click Shuffle first if you’d rather read a different passage.

  • Each passage shows its estimated read time (typically 10–13 seconds) so you know what to expect before you start.
4

Record

Click Start recording and read the passage in a quiet room, speaking clearly, with no background music.

  • These three conditions - clear speech, quiet room, no music - are the product’s own stated guidance, not just general advice, so treat them as requirements rather than suggestions.

Personalize the Voice

Clone voice modal, Step 2 of 3, titled Personalise. A Voice name text field sits above a Style dropdown defaulted to General, which expands to show General, Conversational, Professional, Warm, and Energetic, each with its own icon. A Not happy? Re-record link sits below, with Back and Next buttons at the bottom right.

Step 2: name the voice and pick a style - with a one-click way back to Step 1 if the take needs another pass.
1

Name the Voice

Enter a name in the Voice name field.

  • A specific, memorable name pays off later - “Narrator - Hindi Course” is much faster to find than a generic label once your library has several clones.
2

Set a Style

Choose from General, Conversational, Professional, Warm, or Energetic, or leave it on the default of “General.”

  • This is a separate setting from the personality tags shown in the Voice Library - think of it as a hint for how the clone should be used, not a fixed label the system assigns.
3

Re-record if Needed

Not happy with the take? Click Not happy? Re-record to go back and record again before finishing.

4

Continue

Click Next to move to the final preview and save step.

Preview and Save

Clone voice modal, Step 3 of 3, titled Preview and save. A playback card shows the voice name and an auto-assigned personality tag, a play button, a waveform with elapsed and total time, and a Re-record link in the top right. Below it, Hear your voice on sample lines lists three playable example lines in different tones, a casual line, a news-anchor-style line, and a storytelling line. Below that, a checked consent checkbox reads I acknowledge and consent to the recording, processing, cloning, and use of my voice for the purpose of generating new content on this platform. Back and Save to library buttons sit at the bottom right.

Step 3: hear your clone on three different sample lines before you commit, and give explicit consent to save it.
1

Preview the Recording

Play back your recording using the waveform player at the top of the step.

  • Still not happy? Click Re-record here - you’re not locked into the take from Step 1.
2

Hear It on Sample Lines

Play the three provided sample lines to hear your cloned voice in different registers - casual, news-anchor, and storytelling.

  • This is the fastest way to judge whether a clone will hold up across the range of content you actually plan to use it for, rather than just judging it on the original recording alone.
4

Save to Library

Click Save to library. Sarvam then builds a voice model from your recording - the “Processing” step from the three-step overview.

Once processing finishes, head to Using a Cloned Voice to put it to work.