Instant voice cloning

Inworld's Instant Voice Cloning creates a personalized voice clone in seconds, from as little as 3 seconds of audio. The longer the sample (up to the 30-second limit), the better the speaker similarity. Also known as zero-shot cloning, it's available to all users through Portal or via API.

For a voice clone that more closely matches the original speaker and is more stable, see Professional Voice Cloning Beta. It uses more audio and an asynchronous training step.

Choose one of two ways to use Instant Voice Cloning:

1. Through Inworld Portal - Open Text-to-Speech > Create Voice > Upload or record audio and follow the guided flow.
2. Via API or SDK - Useful if you want to clone a lot of voices or expose this capability to your users.

Clone a Voice in Portal

Go to Inworld Portal

In Portal, select Text-to-Speech in the sidebar. Click Create Voice, then Upload or record audio under Instant Voice Clone.

Upload or record audio samples

Name your voice and select the language, which should match the audio samples. Select Other if you're cloning in a Tier 2 language.

You can either upload or record audio:

• Upload: Drag and drop or browse to upload 1 audio file. Accepted formats: wav, mp3, webm. Maximum file size is 12MB. Audio samples longer than 30 seconds will be automatically trimmed to 30 seconds.
• Record: Click "Record audio" and record your audio. You can use the suggested scripts to help guide your recording, or use your own script. For best results, record in a quiet place to minimize background noise, avoid mic noise, and speak with a variety of emotions to capture the full range of the voice.

Enable "Remove background noise" if you wish to remove background noise from your audio. Confirm you have the rights to clone the voice, then click "Continue".

Check out our Voice Cloning Best Practices for helpful tips and tricks to improve the quality of your voice clones.

Test your cloned voice

Once voice cloning completes, you'll see the "Try your cloned voice" interface. Enter text in the input field and press play to hear your cloned voice. You can test different phrases to ensure the voice sounds as expected.

If the voice doesn't sound quite right, you can delete the voice and start over, create another voice, or test it in the TTS Playground for more advanced testing options.

The number of custom voices you can have depends on your subscription plan. See Custom voice limit to check your usage.

Use your cloned voice via API

To use the cloned voice via API, copy the voice ID for your cloned voice in TTS Playground. Use that value for the voiceId when making an API call. See our Quickstart to learn how to make your first API call.

Clone a Voice via API

If you want to automate voice cloning (for example, to support creator onboarding at scale), use the Voice Cloning API.

• API reference: Clone a voice
• Node.js SDK: cloneVoice() — clone a voice in a few lines with the @inworld/tts SDK
• Python example: example_voice_clone.py
• JavaScript example: example_voice_clone.js

Custom voice limit

Your plan sets how many custom voices your organization can have. Instant clones, designed voices and voices copied in with a copy token all count toward one limit, shared by every workspace in your organization. See the Pricing page for each plan's limit. Professional voice clones are not counted here: they have their own voice slots.

Check your usage

Call Get custom voice capacity with your API key. It is free, and a standard key with read access to voices works:

curl -s 'https://api.inworld.ai/voices/v1/voices:customVoiceCapacity' \
  -H "Authorization: Basic $INWORLD_API_KEY"

{
  "usedCustomVoices": 37,
  "maxCustomVoices": 500
}

usedCustomVoices is how many custom voices your organization has now, across all its workspaces. maxCustomVoices is your plan's limit plus any add-ons. If maxCustomVoices is absent, no limit applies. To track usage over time, call the endpoint on a schedule and alert when usedCustomVoices nears maxCustomVoices.

To list the voices themselves, call List voices with filter=owned="true". That list covers only the API key's workspace, so its count can be lower than usedCustomVoices.

When you reach the limit

When usedCustomVoices equals maxCustomVoices, requests that create a voice fail with 400 (INVALID_ARGUMENT): Clone a voice, Design a voice, Publish a voice and Redeem a voice copy token. The error names your limit:

You have exceeded your plan's maximum number of custom voices (500). Please upgrade to unlock more.

Clone, design and publish return this text in message. Redeem returns it in a field violation on customVoice.totalCustomVoices. Your existing voices keep working. To create more, delete private voices you no longer use, which frees room right away, or upgrade your plan. A voice shared to the Voice Library can't be deleted until you remove it from the library, and removal has a 2-year notice period, so it doesn't free room soon. On a contract plan, ask your Inworld account team for more.

See it in Portal

Portal has no page that always shows your count. The Instant Voice Clone and Voice Design pages under Voices > Create Voice show a notice once you've used 80% of your custom voices, for example "You've used 410 of 500 custom voices." At the limit, those pages lock and show Voice limit reached. Use the API above to check your usage at any time.

Next Steps

Looking for more tips and tricks? Check out the resources below to get started!