A custom TTS voice reads viewer messages in a voice like yours instead of a stock narrator. This guide covers recordings, permissions, storage, moderation and testing.
Library voices are available with every ControlPlay account. Own Voice, the cloned voice this guide covers, is a separate option that ControlPlay enables by invitation.
How TTS in your own voice works
- Add clean voice recordings. Upload one to 25 MP3, WAV, OGG or WebM files, each up to 10 MB, then choose a name and language.
- Create the clone. ControlPlay sends the recordings to ElevenLabs Instant Voice Cloning, which uses them as reference audio without a separate training step.
- Assign it to a TTS action. Viewers spend available Coins to submit text. ControlPlay checks each message before generating audio.
Each action has its own voice, Coin amount, cooldown, volume and audio route. You can keep it active in DND mode, limit it to selected games and mute your microphone during playback.
Where your voice samples are stored and how deletion works
ElevenLabs stores the recordings used for the clone. Its retention documentation says Instant Voice Cloning retains this audio and is not eligible for Zero Retention Mode. ControlPlay does not permanently keep the original uploads on its servers; it stores voice metadata and a generated library preview.
Deleting the voice removes it from your library and queues an ElevenLabs deletion request. Deleting your account does the same. ControlPlay cannot guarantee when ElevenLabs will complete the deletion; its Privacy Policy applies.
Which voice are you allowed to clone?
Use your own voice whenever possible. To clone someone else's, you need their explicit written permission and all rights required to create and use it. Before uploading, you must state that you have those rights.
A public clip is not permission to clone a celebrity, streamer or teammate. Twitch's Community Guidelines still apply, and the ElevenLabs Prohibited Use Policy forbids unauthorized, deceptive or harmful impersonation.
What stops viewers from abusing your voice?
ControlPlay checks viewer text before generating audio. A word filter catches blocked terms and common evasions, while a semantic check reviews the full message. Blocked text is not synthesized; block mode returns the Coins spent.
The TTS moderation guide explains the filtering system. Viewers cannot access your microphone or recordings. They can only submit text through an action you enabled.
Your voice in Discord and game voice chat
On stream, the clone works like any moderated TTS voice. The ControlPlay virtual microphone can also feed Discord or compatible in-game voice chat. Select and test it in each app; compatibility depends on your setup.
Viewer-written TTS uses non-transferable ControlPlay Coins with no monetary value. Up to three optional beta rewards can convert channel-specific Channel Points into Coins at your chosen ratio; ControlPlay calculates the credit automatically. This creates no revenue or 80/20 allocation.
Direct Bits are never available for free-form TTS or voice-chat output. On eligible Affiliate and Partner channels, viewers can confirm through Twitch a permitted, Extension-defined direct interaction or a selected Coin package. Twitch then credits package Coins to their wallet. Packages cannot bypass direct-Bits restrictions. For both permitted routes, Twitch currently allocates 80% of one US cent per Bit to the eligible streamer and 20% to ControlPlay. Twitch can change the program.
Where ControlPlay fits: viewers submit text with Coins, ControlPlay moderates it, and your chosen voice reads it on stream or through the virtual microphone.
Recording samples that sound like you
Consistency matters more than file count:
- Record in a quiet room without music, game audio or hum.
- Keep the same microphone distance and speaking style.
- Speak naturally in the voice and language you want the clone to reproduce.
- Include one speaker with no alerts or voice chat in the background.
ElevenLabs recommends about one to two minutes of clean audio. Preview the clone before enabling it; rebuild it if it sounds noisy or unlike you.
Set up your custom TTS voice
- Create a ControlPlay account, connect Twitch, install the Windows client and activate the Twitch Extension.
- Configure a library voice first with the Twitch TTS setup guide.
- Ask ControlPlay support about Own Voice, since the clone option is enabled by invitation.
- Upload your recordings, choose the language, read the storage notice and state that you have the right to use the voice.
- Preview the clone, select it for one action and set its Coin amount, cooldown, volume and route.
- Test a harmless message from a second viewer account and make a local OBS recording. Test voice-chat routing separately if needed.
Start with one custom-voice action. Add more only after real messages sound clear and blocked text never plays.