If you have been hitting errors with every free caption tool you try — Caption.Ninja, Web Captioner, Zip Captions — it is not your setup. The free category is collapsing because Chrome is deprecating the Web Speech API every one of them depends on.
No credit card. 6-hour free trial. Works on any browser.
The same root cause: Chrome's Web Speech API is being deprecated. Google previously offered free server-side speech recognition through the browser. That subsidized every free caption tool on the market.
Google is now migrating to on-device SODA (Speech On-Device API) models. During the migration, the old server-side service is being throttled, and the new on-device service is not reliably shipping to all users. Multiple Chromium bug reports document the breakage:
| Tool | Status | Reason |
|---|---|---|
| Web Captioner | Shut down Oct 31, 2023 | Maintainer abandoned + API deprecation |
| Caption.Ninja | Failing in Chrome | Web Speech API deprecation |
| Zip Captions | Same issues | Same Web Speech API dependency |
| DIY browser-source scripts | Breaking | Same API |
| LocalVocal (OBS plugin) | Works (different stack) | Uses local Whisper, not Web Speech API. But heavy GPU. |
| StreamTranslate | Works | Deepgram server-side, no Web Speech API dependency |
LocalVocal — runs Whisper locally on your GPU. Truly free, no API costs, no Web Speech API dependency. Downsides: heavy GPU usage (5-20% during streams), breaks on OBS plugin updates, Linux/Windows/Mac install required, single mic only, no real-time translation race.
If you have GPU headroom and want truly free, LocalVocal is currently the most viable free option. If you do not want to install a plugin + manage GPU + restart-on-OBS-update, you need a managed solution.
Tools using paid speech recognition APIs are not affected by Chrome's Web Speech API deprecation:
Of these, StreamTranslate is the only one purpose-built for streamers with browser-source OBS integration, sub-500ms latency, and entry pricing at $9.99 one-time. 6-hour free trial, no card.
Browser-native speech recognition was always a Google subsidy. Chrome would handle the speech recognition for free, and any web developer could build a captioning tool on top. That model is ending. Google's on-device SODA approach moves the cost to the user's machine (storage + CPU), and reliability is much lower while the migration completes.
The free captioning category is going to look very different in 2027. Either tools migrate to paid speech recognition APIs and charge users, or they accept unreliable browser-native speech recognition with no SLA.
No credit card. One URL pasted into OBS. Live in 60 seconds.
Launch StreamTranslate →