Transparency
Voices & credits
Last updated: 24 September 2026
YourAIVoice is built entirely from openly licensed models and software. We credit them here, as their licences require. YourAIVoice is an independent product of Markaroo Solutions and is not affiliated with or endorsed by any of the projects listed below.
The voices
All voices come from Kokoro-82M by hexgrad, an open-weight speech model released under the Apache License 2.0. The names and quality grades below are the ones the Kokoro authors publish in their voice list. We only offer voices graded C or better, plus the ungraded Spanish and Portuguese voices. Kokoro's Japanese and Chinese voices are left out for now, because they need extra pronunciation software to sound right.
The Telugu voices are the exception: Kokoro has no Telugu, so they are generated on your device by eSpeak NG (GPL-3.0), a free open-source synthesizer. eSpeak produces correct but robotic-sounding Telugu. Their names (Ananya, Lalita, Teja) are ours.
| Language | Voice | Gender | Official grade | Voice ID |
|---|---|---|---|---|
| English (US) | Heart | Female | A | af_heart |
| English (US) | Bella | Female | A- | af_bella |
| English (US) | Nicole | Female | B- | af_nicole |
| English (US) | Sarah | Female | C+ | af_sarah |
| English (US) | Kore | Female | C+ | af_kore |
| English (US) | Michael | Male | C+ | am_michael |
| English (US) | Fenrir | Male | C+ | am_fenrir |
| English (US) | Puck | Male | C+ | am_puck |
| English (UK) | Emma | Female | B- | bf_emma |
| English (UK) | Isabella | Female | C | bf_isabella |
| English (UK) | George | Male | C | bm_george |
| English (UK) | Fable | Male | C | bm_fable |
| हिन्दी · Hindi | Alpha | Female | C | hf_alpha |
| हिन्दी · Hindi | Beta | Female | C | hf_beta |
| हिन्दी · Hindi | Omega | Male | C | hm_omega |
| हिन्दी · Hindi | Psi | Male | C | hm_psi |
| Español · Spanish | Dora | Female | Not graded | ef_dora |
| Español · Spanish | Alex | Male | Not graded | em_alex |
| Français · French | Siwis | Female | B- | ff_siwis — Trained partly on the SIWIS database (CC BY 4.0) |
| Italiano · Italian | Sara | Female | C | if_sara |
| Italiano · Italian | Nicola | Male | C | im_nicola |
| Português (Brasil) | Dora | Female | Not graded | pf_dora |
| Português (Brasil) | Alex | Male | Not graded | pm_alex |
Training-data attributions
According to its authors, Kokoro-82M was trained on permissively licensed and public-domain audio. Two of its training datasets are licensed under Creative Commons Attribution, and are credited here:
- The SIWIS French Speech Synthesis Database by Pierre-Edouard Honnet, Alexandros Lazaridis, Philip N. Garner and Junichi Yamagishi (University of Edinburgh DataShare), licensed under CC BY 4.0. Kokoro's authors used part of it to train the model.
- Koniwa (github.com/koniwa/koniwa), licensed under CC BY 3.0. Kokoro's authors used a small part of it to train the model.
Open-source software
The voice engine runs in your browser, so these components are delivered to your device from our website:
| Component | What it does | Licence |
|---|---|---|
| Kokoro-82M (int8 ONNX file from kokoro-onnx) | Speech model and voices | Apache-2.0 (model) · MIT (kokoro-onnx) |
| ONNX Runtime Web | Runs the model in your browser | MIT |
| eSpeak NG (rhasspy fork) | Turns words into pronunciations | GPL-3.0-or-later (licence text) |
| espeak-ng (WebAssembly build) | Generates the Telugu voices on your device | GPL-3.0-or-later (licence text) |
| piper-phonemize, WebAssembly build by @diffusionstudio/piper-wasm | Runs eSpeak NG in the browser | MIT (piper-phonemize) · includes eSpeak NG (GPL-3.0-or-later) |
| lamejs (@breezystack/lamejs), a JavaScript port of LAME | MP3 encoding | LGPL-3.0 (licence notes) |
| Plus Jakarta Sans | Font (served from our own website) | SIL OFL 1.1 (licence text) |
We did not change the source code of these components. eSpeak NG is used in two ways: a phonemizer (via piper-phonemize) that converts words to pronunciations for the Kokoro voices, and a full eSpeak NG build that generates the Telugu voices directly, including the Telugu pronunciation data. The GPL and LGPL components are provided as separate files (/assets/engine/), and their complete source code is available at the links above. The voice samples on this website were made with the same Kokoro-82M voices.
Arabic voices (optional)
If Arabic (UAE) voices are offered, they are provided by Microsoft Azure AI Speech under Microsoft's terms. Those voices are not open source and are used through a paid Microsoft service.
Responsible use
These voices are AI-generated. Please don't use them to impersonate anyone or to mislead listeners, and say that a voice is AI-generated where that matters. See our Terms & Conditions.