Kokoro provides an inference library for its open-weight 82-million-parameter text-to-speech model.
Open source
22 alternativesKokoro-82M
The Kokoro-82M model provides open-weight text-to-speech with 82 million parameters.
About Kokoro-82M
Kokoro-82M is an open-weight text-to-speech model with 82 million parameters, published as a model page on Hugging Face. It turns written text into spoken audio and is described as lightweight yet comparable in quality to larger models, while being faster and more cost-efficient. Its weights are Apache-licensed, and the page states it can be deployed from production environments to personal projects. It uses the misaki grapheme-to-phoneme library under the hood.
Key features
- Text-to-speech model with 82 million parameters
- Apache-licensed weights for deployment in projects and commercial APIs
- Version 1.0 covers 8 languages and 54 voices
- Python usage via the kokoro package with soundfile and espeak-ng
- Decoder-only architecture; no diffusion and no encoder release
- Trained on permissive or non-copyrighted audio plus IPA phoneme labels
Best for: Developers and researchers who want a small, Apache-licensed TTS model they can run themselves or access through third-party APIs.
Good for
- Apache-licensed weights permit deployment in production or personal projects
- Lightweight 82M-parameter design described as faster and more cost-efficient than larger models
- Training used permissive or non-copyrighted audio and IPA phoneme labels
Watch out for
- Decoder-only: no encoder is released and no diffusion is used
- The model page warns that sites with 'kokoro' in the root domain are not affiliated and may be scams
- Version 0.19 was trained on under 100 hours of audio, and v1.0 on a few hundred hours
Kokoro-82M at a glance
- Category
- AI & Machine Learning › AI Voice & Audio
- Type
- Online tool
- Pricing
- Not confirmed. Check the site
- Platforms
- Web
- Open source
- Not stated
- Last checked
- Oct 5, 2026 · Online when checked
- Listed
- Oct 5, 2026
Kokoro-82M alternatives
Tools and sites you could use instead, ranked by how closely they match.
This Kokoro TTS demo converts English text into speech with selectable voices and speaking speed.
CassetteAI provides music, sound-effect, and text-to-speech models through an SDK designed to run on edge hardware.
Inworld provides real-time text-to-speech, speech-to-text, and language model inference through modular APIs.
NeMo Parakeet ASR Models attain strong speech recognition accuracy while being efficient for inference. Available in CTC and RNN-Transducer variants.
UnificAlly provides one API for generating video, images, music, and speech across more than 100 AI models.
Popular in AI Voice & Audio
Well-known listings from the same topic.
Google Illuminateilluminate.google.comAlteredaltered.aiELSA Speakelsaspeak.comKreadoAIkreadoai.comPodwisepodwise.aiAddsubtitleaddsubtitle.ai
Kokoro-82M FAQ
What licence applies to the weights?
The model page states the weights are Apache-licensed and that the model has been deployed in numerous projects and commercial APIs.
Does the model page run an official website at a 'kokoro' domain?
No. The page states that any website containing 'kokoro' in its root domain is not owned by and not affiliated with the model page or its author.
How many languages and voices does the current release cover?
Release v1.0, published 27 January 2025, is listed with 8 languages and 54 voices.
Is something wrong or out of date?
Tell us and an editor will check it.
Own this website?
Claim this listing
Claim the listing and suggest updates yourself. An editor checks every claim first.