Kokoro TTS
github.com
- Free plan
- Windows
- macOS
- Linux
- Web
- API
- Self-hosted
What Kokoro TTS is
Kokoro TTS is a free, open-weight text-to-speech model with 82 million parameters and Apache-licensed weights. It uses a decoder-only architecture with StyleTTS 2 and ISTFTNet and supports eight languages. The maker describes it as lightweight and more efficient than larger models while offering comparable quality. A Python pipeline example lets users select a voice and set speech speed; generated audio is 24 kHz and can be saved as WAV. The kokoro.js library runs the model locally in a browser through Transformers.js, with Node CPU use also documented, and can stream speech as text is added. The model is presented for personal or production projects, and commercial use is permitted. Python setup uses the kokoro package, soundfile, and espeak-ng, with Misaki handling grapheme-to-phoneme processing. The README includes Windows guidance for installing espeak-ng and Apple Silicon GPU instructions for macOS M1 through M4.
Who it suits
Kokoro TTS may suit developers who want to generate speech from Python or run a model locally in a browser. Its Apache-licensed weights and stated commercial-use permission also fit personal and production projects.
Good to know, and what to watch
- Free, Apache-licensed model weights.
- Supports eight languages.
- Python pipeline supports voice and speed selection.
- Browser library can stream speech as text is added.
- Commercial use is permitted.
- WAV is the listed audio export format.
- Python setup uses kokoro, soundfile, and espeak-ng.
- The model is described as having 82 million parameters.
Plans
Every plan and priceKokoro open-weight model
Free plan
Getting started
- Open the Kokoro repository website
- Install the kokoro package, soundfile and espeak-ng for Python use
- Choose a voice and speech speed in the Python pipeline
- Generate 24 kHz speech and save it as WAV
- Run kokoro.js locally in a browser through Transformers.js or use documented Node CPU support
Limits to check
The model supports eight languages and exports WAV audio. Voice cloning is not available.
The Club's review
Kokoro TTS is a free, open-weight speech model for developers and technically confident teams that want to run text-to-speech under their own control. It is most compelling when licensing, local inference and deployment efficiency matter more than turnkey voice-production features.
Overview
Kokoro uses 82 million parameters in a decoder-only design built with StyleTTS 2 and ISTFTNet. Its maker positions it as lighter, faster and more cost-efficient than larger models while aiming for comparable quality. That makes it a sensible candidate for applications where inference resources and operating cost need to stay modest, although teams still have to assemble and maintain the runtime.
The weights use the Apache license and are intended for personal or production deployment, with commercial use supported. Training used permissive or non-copyrighted audio and IPA phoneme labels from a dataset of a few hundred audio hours. There is no voice cloning, so projects requiring a particular speaker's identity should look elsewhere.
Key features
- Python controls and WAV generation: The documented pipeline lets developers choose a voice and set speech speed. It renders audio at 24 kHz and saves WAV files, which suits straightforward processing pipelines but gives you only the export format documented for the model.
- Local browser and Node paths: The kokoro.js library runs inference in a browser through Transformers.js and documents CPU use in Node. This can keep speech processing close to an application and avoid a hosted inference dependency, provided the target environment can carry the model and its runtime.
- Streaming JavaScript output: kokoro.js documents generating speech as text arrives. That is useful for conversational or progressively rendered interfaces where waiting for a complete passage would make the experience feel slow.
- Eight languages: Kokoro supports eight languages, making it more practical for multilingual products than a single-language engine, while still requiring you to verify that its language coverage matches your audience.
- Developer examples: Examples cover Google Colab, a Hugging Face demo, inference providers and notebooks. They provide several starting points for experimentation, but the core package remains an engineering-oriented workflow rather than a finished desktop editor.
- Phoneme and installation stack: Python setup uses the kokoro package, soundfile and espeak-ng, with Misaki handling grapheme-to-phoneme processing. The README includes Windows espeak-ng guidance and Apple Silicon GPU instructions for macOS M1 through M4, useful details for reducing platform friction.
- Community support: A Discord community connects Kokoro users and contributors, giving teams a place to discuss implementation alongside the repository examples.
Pricing
Kokoro open-weight model costs free. The plan includes the 82-million-parameter model and Apache-licensed weights. There are no paid tiers, seat charges or subscription terms stated, so it fits developers, researchers and commercial teams that prefer to operate their own infrastructure. The trade-off is that you take responsibility for dependencies, deployment and the audio pipeline instead of receiving a managed service.
Platforms
Kokoro is offered for API, Linux, macOS, self-hosted deployments, the web and Windows. Python provides the primary documented model workflow, while JavaScript supports local browser execution and Node CPU use. macOS users on Apple Silicon have specific acceleration guidance, and Windows users have installation instructions for espeak-ng. These options favor teams comfortable choosing and maintaining their execution environment.
Who it's for
Kokoro suits developers building speech into products, internal tools or automations that need commercial-use rights, local execution and an open model. It is also a good fit for cost-conscious projects that can support Python or JavaScript dependencies and accept WAV output. It is a weaker choice for nontechnical creators seeking a polished voice studio, for teams that need voice cloning, or for workflows requiring a broad range of export formats without additional conversion work.
Pros and cons
Pros
- Free Apache-licensed weights: Personal and production deployment, including commercial use, is supported without a recurring plan.
- Multiple local routes: Python, browser JavaScript and Node documentation give teams alternatives to a single hosted endpoint.
- Efficient model size: The 82-million-parameter footprint is aimed at lower resource and cost demands than larger models.
- Useful runtime controls: Voice selection, speed adjustment and documented streaming cover common application needs.
Cons
- Setup is your responsibility: kokoro, soundfile, espeak-ng and Misaki add environment work before speech generation.
- WAV is the documented export: Projects needing other delivery formats must handle conversion themselves.
- No voice cloning: It cannot reproduce a specific speaker for branded or character-led workflows.
- Technical orientation: Examples, notebooks and inference libraries favor builders over users looking for a ready-made editing interface.
Alternatives
Typecast is the better fit for a hosted, cross-device service with Android, iOS, web and API access, a free tier with three projects and 3,000 lifetime credits, and a free trial.
Google Cloud Text-to-Speech suits teams that want metered cloud delivery: standard voices include 0 to 4 million characters free each month, with usage billed per character beyond that allowance.
LOVO AI is an alternative for a freemium web/API workflow with five projects on its free plan and a Pro+ option offering 20 voice-generation hours per month and unlimited projects.
Fish Audio offers a freemium route across the same broad desktop, web, API and self-hosted platform set, with a free tier capped at 8,000 credits monthly, seven minutes per generation and 500 characters per generation.
Speechify is aimed at users who want a consumer-facing service, with 10 voices and 1.5x speed on free, and a Premium plan at $29 a month.
Voicemaker provides a hosted API/web option with a free plan limited to 250 characters per conversion and personal use, while its Starter plan costs $5 a month.
Cartesia Sonic is a metered API/web alternative with 20K credits per month on free and 100K credits per month on its $5-per-month Pro plan.
Chatterbox TTS is another free, self-hosted open-source option under the MIT license for Linux and macOS.
For broader discovery, browse AI Voice Generators, Game Voice Generators and Text-to-Speech Software.
Verdict
Choose Kokoro TTS if you need a free, commercially usable open model with Python and JavaScript deployment paths and can own the setup. Its main advantage is control without a subscription; look elsewhere if you need voice cloning, richer export choices or a managed interface that hides the inference environment.
The verdict
Kokoro TTS provides a free model with Python and JavaScript usage paths, including local browser execution. Check the documented setup dependencies and WAV output format against your intended workflow.
Questions people ask
Is Kokoro TTS free?
Yes. The listed Kokoro open-weight model plan costs nothing, and a free plan is available.
Can it be used commercially?
Yes. Commercial use is listed as permitted, and the weights are Apache-licensed.
Which platforms are supported?
Kokoro is listed for API, Linux, macOS, self-hosted, web and Windows use.
Does Kokoro offer voice cloning?
No. Voice cloning is listed as unsupported.
What audio format does it export?
The listed export format is WAV; the Python example generates 24 kHz audio.
What the maker publishes
- API access
- Yes source
- Architecture
- The model uses StyleTTS 2 and ISTFTNet and is decoder-only. source
- Audio output
- The Python example generates audio at 24 kHz and saves it as WAV files. source
- Commercial use
- Yes source
- Dependencies
- The documented Python installation uses the kokoro package, soundfile, and espeak-ng; Kokoro uses Misaki for grapheme-to-phoneme processing. source
- Efficiency
- The maker describes Kokoro as lightweight, faster, and more cost-efficient than larger models while delivering comparable quality. source
- Export formats
- WAV source
- Free plan
- Yes source
- Integrations
- The maker provides examples using Google Colab and points to a Hugging Face demo; the model page also lists inference providers and notebooks. source
- JavaScript library
- The kokoro.js library runs the model locally in a browser through Transformers.js and also documents Node CPU use. source
- Languages
- 8 source
- Languages
- The usage documentation lists American English, British English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, and Mandarin Chinese. source
- License
- The model weights are Apache-licensed, and the maker says Kokoro can be deployed in production or personal projects. source
- Platform guidance
- The README gives Windows installation guidance for espeak-ng and Apple Silicon GPU acceleration instructions for macOS M1 through M4. source
- Platforms
- web,windows,macos,linux,api,self_hosted source
- Purpose
- Kokoro is an open-weight text-to-speech model with 82 million parameters. source
- Streaming
- The JavaScript library documents streaming generated speech as text is added over time. source
- Support
- The repository links to a Discord community for Kokoro users and contributors. source
- Training data
- The model card says Kokoro was trained on permissive or non-copyrighted audio data and IPA phoneme labels, with a dataset of a few hundred audio hours. source
- Voice and speed controls
- The Python example selects a voice and sets speech speed through the pipeline. source
- Voice cloning
- No source
Checked against the maker's pages on 5 Oct 2026.