Text to SpeechQuery Akuluakulu Muzika Kusintha mawu kukhala malemba Chizindikiro cha mawu Voice Design Marketplace Zambiri

Tool

Kulembetsa kwaulere

Help Build TTS for Every Language

We only need 2 hours of recordings per language to train a TTS model. Contribute your voice to help create free, open-source text-to-speech for all 100 languages.

32 languages supported 68 languages need your help
1

Record

Read sentences aloud in your language using your microphone. Each recording takes 5-15 seconds.

2

We Train

Once we collect 2+ hours of quality recordings per language, we fine-tune a TTS model on the data. Models are published open-source.

3

Everyone Benefits

The trained model becomes available on TTS.ai for free. Contributors get credits and early access.

Language Progress

Afrikaans
0.0h / 20h 0 helpers
Albanian
0.0h / 20h 0 helpers
Amharic
0.0h / 20h 0 helpers
Armenian
0.0h / 20h 0 helpers
Assamese
0.0h / 20h 0 helpers
Azerbaijani
0.0h / 20h 0 helpers
Basque
0.0h / 20h 0 helpers
Belarusian
0.0h / 20h 0 helpers
Bengali
0.0h / 20h 0 helpers
Bosnian
0.0h / 20h 0 helpers
Bulgarian
0.0h / 20h 0 helpers
Burmese
0.0h / 20h 0 helpers
Catalan
0.0h / 20h 0 helpers
Chichewa
0.0h / 20h 0 helpers
Croatian
0.0h / 20h 0 helpers
Esperanto
0.0h / 20h 0 helpers
Estonian
0.0h / 20h 0 helpers
Filipino
0.0h / 20h 0 helpers
Galician
0.0h / 20h 0 helpers
Gujarati
0.0h / 20h 0 helpers
Haitian Creole
0.0h / 20h 0 helpers
Hausa
0.0h / 20h 0 helpers
Hebrew
0.0h / 20h 0 helpers
Igbo
0.0h / 20h 0 helpers
Indonesian
0.0h / 20h 0 helpers
Irish
0.0h / 20h 0 helpers
Javanese
0.0h / 20h 0 helpers
Kannada
0.0h / 20h 0 helpers
Khmer
0.0h / 20h 0 helpers
Kinyarwanda
0.0h / 20h 0 helpers
Kurdish
0.0h / 20h 0 helpers
Kyrgyz
0.0h / 20h 0 helpers
Lao
0.0h / 20h 0 helpers
Latvian
0.0h / 20h 0 helpers
Lithuanian
0.0h / 20h 0 helpers
Luxembourgish
0.0h / 20h 0 helpers
Macedonian
0.0h / 20h 0 helpers
Malagasy
0.0h / 20h 0 helpers
Malay
0.0h / 20h 0 helpers
Malayalam
0.0h / 20h 0 helpers
Maltese
0.0h / 20h 0 helpers
Maori
0.0h / 20h 0 helpers
Marathi
0.0h / 20h 0 helpers
Mongolian
0.0h / 20h 0 helpers
Odia
0.0h / 20h 0 helpers
Pashto
0.0h / 20h 0 helpers
Persian
0.0h / 20h 0 helpers
Punjabi
0.0h / 20h 0 helpers
Shona
0.0h / 20h 0 helpers
Sindhi
0.0h / 20h 0 helpers
Sinhala
0.0h / 20h 0 helpers
Slovenian
0.0h / 20h 0 helpers
Somali
0.0h / 20h 0 helpers
Sundanese
0.0h / 20h 0 helpers
Tajik
0.0h / 20h 0 helpers
Tamil
0.0h / 20h 0 helpers
Tatar
0.0h / 20h 0 helpers
Telugu
0.0h / 20h 0 helpers
Thai
0.0h / 20h 0 helpers
Turkmen
0.0h / 20h 0 helpers
Urdu
0.0h / 20h 0 helpers
Uyghur
0.0h / 20h 0 helpers
Uzbek
0.0h / 20h 0 helpers
Welsh
0.0h / 20h 0 helpers
Xhosa
0.0h / 20h 0 helpers
Yiddish
0.0h / 20h 0 helpers
Yoruba
0.0h / 20h 0 helpers
Zulu
0.0h / 20h 0 helpers
Arabic
TTS Available
Chinese
TTS Available
Czech
TTS Available
Danish
TTS Available
Dutch
TTS Available
English
TTS Available
Finnish
TTS Available
French
TTS Available
Georgian
TTS Available
German
TTS Available
Greek
TTS Available
Hindi
TTS Available
Hungarian
TTS Available
Icelandic
TTS Available
Italian
TTS Available
Japanese
TTS Available
Kazakh
TTS Available
Korean
TTS Available
Nepali
TTS Available
Norwegian
TTS Available
Polish
TTS Available
Portuguese
TTS Available
Romanian
TTS Available
Russian
TTS Available
Serbian
TTS Available
Slovak
TTS Available
Spanish
TTS Available
Swahili
TTS Available
Swedish
TTS Available
Turkish
TTS Available
Ukrainian
TTS Available
Vietnamese
TTS Available

Contribute a Recording

Create a free account to start contributing your voice. Your recordings help build free TTS models for everyone.

Sign Up to Contribute

Already Speak a Supported Language?

You can still help by sharing your voice for others to use. Upload a voice recording and let the community clone your voice for their TTS projects.

Share Your Voice

Funso Lofunsidwa Kawirikawiri

The voice contribution program lets you donate short voice recordings to help build free, open-source TTS models for languages that currently lack support. Your recordings help train AI models that anyone can use.

Select your language, then read sentences aloud using your microphone. Each recording is 5-15 seconds long. You can contribute as few or as many recordings as you like. All accepted recordings go toward training a free TTS model for your language.

Accepted recordings are used exclusively to train open-source TTS models. The trained models are released under Apache 2.0 license on GitHub and HuggingFace. Your recordings are not sold or shared with third parties. Contributors are credited in the model release.

We currently support 32 languages with TTS models. The remaining 68 languages on our site need community voice data. The contribution page shows a progress bar for each language — languages closer to their target hours are prioritized for training.

We target 20 hours of recordings per language for a good single-speaker model. With 100+ contributors each recording 10-15 minutes, we can reach this goal. Even small contributions help — every recording brings us closer.

Yes! Contributors earn TTS credits for accepted recordings. Top contributors get a "Voice Pioneer" badge and are credited on the trained model. You also get early access to your language TTS model before public release.

Yes! More data improves quality. Contributions for supported languages help us build multi-speaker models with more voice variety, better accents, and improved naturalness.

A quiet room and any microphone — even a phone or laptop mic works. For best quality, use a headset or USB microphone. Avoid background noise, echo, and multiple speakers. The system will check your audio quality before accepting.

Each recording goes through automated quality checks: signal-to-noise ratio (SNR > 20dB), Whisper transcript matching against the expected sentence, duration within target range, and single-speaker detection. Recordings that fail can be re-recorded.

You can preview the contribution interface without an account, but recordings are linked to your TTS.ai account so we can credit you properly and track your contributions. Signing up is free and takes 30 seconds.

When a language accumulates enough data (typically 20+ hours), our team begins model training. Training takes 1-2 weeks on our GPU cluster. The resulting model is tested, published on HuggingFace under Apache 2.0, and added to TTS.ai for everyone to use.

Yes. The contribution dashboard shows a progress bar for every language, including total recordings, hours collected, number of contributors, and percentage toward the training target. You can also see your personal contribution stats.
5.0/5 (1)

Ndinu okonzeka kuyamba?

Sign up for free and get 50 credits. No credit card required.

Kulembetsa kwaulere Onani mtengo