Kokoro
Open-weight text-to-speech model with 82 million parameters that you run yourself, under the Apache-2.0 license.
AI Audio & VoiceFreeverified Listed on Open source
Is Kokoro free?
Yes, Kokoro has a free plan. What you get: Open source (Apache-2.0 license): free model weights you run yourself with Python. To get it: install the kokoro library with pip, plus espeak-ng; then create a pipeline with a language code and a voice; then generate audio from your text and save it as WAV files. Verified on huggingface.co.
Free plan at a glance
- Includes
- Open source (Apache-2.0 license): free model weights you run yourself with Python
Open weights under the Apache-2.0 license that run on your own machine; the maker's pages state no usage limit.
- Duration
- Free forever
- Commercial use
- Allowed on the free plan
How to get it free
Self-host- Install the kokoro library with pip, plus espeak-ng
- Create a pipeline with a language code and a voice
- Generate audio from your text and save it as WAV files
Good to know
- Needs espeak-ng for English fallback and some non-English languages
- Non-English voices may be thin; French has one voice
github.com, Checked Source
Overview
Kokoro is a small open-weight text-to-speech model released under the Apache-2.0 license, so it can be used in personal and production projects. You run it with the kokoro Python library, with voices in English, Japanese, Mandarin Chinese, Hindi, Spanish and other languages.
Key features
- 82-million-parameter model that the maker says matches larger models while running faster
- Voices for American and British English, Japanese, Mandarin Chinese, Spanish, French, Hindi, Italian and Brazilian Portuguese
- Python library installed with pip, with adjustable speed and WAV output
- GPU acceleration on Apple Silicon Macs through an environment variable
- Weights licensed for commercial use, already used in projects and commercial APIs
Best for
- People
- Developers adding speech to apps and products, People who want text to speech that runs on their own computer, Teams that need a speech model licensed for commercial use
Limitations
- Runs from Python code: you install the kokoro library and espeak-ng yourself
- The maker says non-English support may be thin, and French has a single voice
- Voices can sound weak on very short lines and rushed on very long ones
Screenshot

FAQ
- Is Kokoro free?
- Yes. Kokoro has a free plan: Open source (Apache-2.0 license): free model weights you run yourself with Python. We last checked huggingface.co's pricing page on 2026-10-24.
- What does the Kokoro free plan include?
- Open source (Apache-2.0 license): free model weights you run yourself with Python. Source: https://huggingface.co/hexgrad/Kokoro-82M (checked 2026-10-24).
- How do I get Kokoro for free?
- Steps: Install the kokoro library with pip, plus espeak-ng; Create a pipeline with a language code and a voice; Generate audio from your text and save it as WAV files. Start at https://github.com/hexgrad/kokoro (checked 2026-10-24).
- Do I need an account to use Kokoro for free?
- Not for the free option: the free route is to self-host Kokoro (Install the kokoro library with pip, plus espeak-ng). Checked 2026-10-24.
- What's the catch with Kokoro's free plan?
- Needs espeak-ng for English fallback and some non-English languages. Non-English voices may be thin; French has one voice. Checked 2026-10-24.
- What are free alternatives to Kokoro?
- NaturalReader, Fish Audio and Kitten-TTS all have free plans. We checked their free offers between 2026-10-24 and 2026-10-25.
- Can I use Kokoro commercially?
- Yes. The model card says it is an Apache-licensed model that has been deployed in numerous projects and commercial APIs, and the maker welcomes its use in real use cases.Source: huggingface.co/hexgrad/Kokoro-82M
- Which languages does Kokoro support?
- Its voices cover American and British English, Japanese, Mandarin Chinese, Spanish, French, Hindi, Italian and Brazilian Portuguese. The voices page warns that non-English support may be thin, and French has one voice.Source: huggingface.co/hexgrad/Kokoro-82M/blob/main/VOICES.md
- Are websites with "kokoro" in their name official?
- No. The model card says any website containing "kokoro" in its root domain is not owned by or affiliated with the model page or its author.Source: huggingface.co/hexgrad/Kokoro-82M
Free alternatives
NaturalReader
Text-to-speech reader for documents and web pages.
Free forever, Free voices with daily limits

Fish Audio
Text-to-speech and voice cloning.
Free forever, Limited monthly generations

Kitten-TTS
Open-source text-to-speech models with voice cloning that run in real time on a CPU, with no GPU needed.
Free forever, Run the models yourself; hosted API has 30 free minutes

TTSMaker
Free online text-to-speech tool with many voices.
Free forever, Free text-to-speech with weekly character limits

Transkriptor
AI transcription.
Free forever, 'Free - 90 Minutes' of transcription (paid plans from Lite $9.99/month)
Compare
- Kokoro vs ElevenLabs · free plans side by side
Sources
- Free offer
- Verified on huggingface.co/hexgrad/Kokoro-82M
…is an Apache-licensed model, and Kokoro has been deployed in numerous projects and commercial…
Added by our team. We check that a free option exists and record the date; signing up is between you and Kokoro, so read their privacy policy and terms first.
Is this your tool? Claim it · Promote it