Our take
Inactiveno commits in over a year- Activity
- 20
- Community
- 19
- Issues
- 21
- Pull requests
- 21
+0 stars in the last 30 days
- Fewer contributors than 86% of the projects we track
This is a focused option for private, browser-based Kokoro speech generation, but its inactive status and highly concentrated contributor activity make it a less certain choice for teams needing ongoing maintenance.
Good fit if
- You want to generate speech locally and avoid sending text to a server.
- Your users can run a modern browser and accept model loading and browser-side processing.
- You are comfortable evaluating an inactive project with most contributions from one person.
Look elsewhere if
- You need evidence of recent maintenance or a steady release history.
- You need a server-side or API-based text-to-speech integration, which this project does not provide.
- Firefox support is important, since the README says it does not currently seem to work.
All health signals
| Last commit | 2025-06-12 (1.3 years ago) |
|---|---|
| Commits, last 90 days | 0 |
| Releases, last 12 months | 0 |
| Contributors | 2 (top contributor: 91% of commits) |
| Issues closed, last 90 days | 0 |
| Pull requests merged, last 90 days | 0 |
| Project age | 17 months |
Checked on 2026-10-10 with the GitHub API.
Project history
Daily GitHub snapshots since Oct 7, 2026. History covers stars and forks; health signals are not tracked historically.
Stars
372Forks
39Change since Nov 8, 2026: -2 stars·0 forks
Overview
StreamingKokoroJS is a web app for generating speech from text with the Kokoro-82M model. It processes text in the browser rather than sending it to a server, addressing privacy concerns for people who want local text-to-speech.
Audio is generated in chunks and can be played as it streams or downloaded. The project is aimed at people comfortable using a modern browser and loading a sizable model on their device.
Key Features
- Runs text-to-speech locally in the browser, without sending text to a server.
- Streams audio as text is processed in chunks.
- Uses WebGPU when available and falls back to WebAssembly.
- Runs generation in Web Workers to keep the interface responsive.
- Supports multiple voice styles for different languages.
- Lets users download generated audio.
- Can be configured to load a local model instead of downloading one from Hugging Face.
Use Cases
- Privacy-conscious users can generate speech from personal text without uploading it to a service.
- People who want to listen to long passages can play speech as it is generated rather than wait for a complete file.
- Developers exploring browser-based machine learning can use it as a working example of Kokoro-JS and Transformers.js.
- Users who need an audio file can generate speech and download it from the browser.
What you need
Detected in the repository
- JavaScript (main language on GitHub)
License in plain words
Apache-2.0permissive
- Commercial use: yes
- Modify and redistribute: yes
- You must keep: the license, the NOTICE file and a note of your changes
- Share your changes: no
- Includes an explicit patent grant from the contributors:
A summary, not legal advice: the LICENSE file is what applies.
Getting Started
Clone the repository, serve its files with a local web server, then open the app in a modern browser. See the README for setup details and the online demo.
Alternatives
- ChatTTS: ChatTTS is a dialogue-focused model for Chinese and English, rather than a browser app built around streaming Kokoro speech.
- GPT-SoVITS: GPT-SoVITS is a Python toolkit for voice cloning and synthesis, rather than an in-browser, on-device Kokoro app.
- csm: CSM generates speech from text and conversation context but requires CUDA-compatible hardware and Hugging Face checkpoints.
| Project | Language | License | Stars | Status |
|---|---|---|---|---|
| StreamingKokoroJS | JavaScript | Apache-2.0 | 372 | Inactive |
| ChatTTS | Python | AGPL-3.0 | 39.9k | Quiet |
| GPT-SoVITS | Python | MIT | 62.7k | Active |
| csm | Python | Apache-2.0 | 14.7k | Inactive |
Considerations
The repository has had no commits in over a year and no releases in the past year. Most code contributions come from one person, and the measured activity shows no recent issue closures or merged pull requests, so teams should plan to assess and maintain it themselves if needed.
The model is about 300 MB and is cached after its first load, which can mean a substantial initial download. WebGPU can accelerate generation on compatible browsers, but the project also provides a WebAssembly fallback; Firefox is noted as not working at the moment. The README describes local-model configuration, but does not document a packaged server-side deployment or API.
Found this useful?
Share it with someone who would like StreamingKokoroJS.