🎯 Try StreamTranslate free for your next stream — 60-second setup, no card requiredStart Free Trial →
⚠ Current issue · May 2026

Caption.Ninja Speech Recognition Failing in Chrome?

If you're seeing "Speech recognition service is currently unavailable in Google Chrome. Try using Microsoft Edge instead" — you're not alone. Here's the technical reason it's happening, why Edge isn't a real fix, and what to use instead.

⚠ What you're seeing
A red banner on Caption.Ninja saying Speech recognition service is currently unavailable in Google Chrome. Try using Microsoft Edge instead. Or a popup warning that "the service will fail completely if more than one transcription session is currently active on your computer." These errors are increasing as Google deprecates the Web Speech API.
Try StreamTranslate Free → View Pricing

No credit card. 6-hour free trial. Works on any browser — Chrome, Safari, Firefox, Edge.

TL;DR
Caption.Ninja uses Chrome's Web Speech API for free speech-to-text. Google is deprecating that API and migrating to on-device SODA models, causing intermittent failures across Caption.Ninja, Web Captioner, and any other tool that relies on it. Switching to Edge sometimes helps but isn't a fix — Edge has the same underlying issues. The architectural problem is that browser-native speech recognition is no longer reliable for streaming. StreamTranslate uses dedicated Deepgram speech recognition that runs server-side, doesn't depend on Chrome, and works in every browser. 6-hour free trial, no credit card.

Why is this happening?

Caption.Ninja is a free, open-source captioning tool built on top of Chrome's built-in Web Speech API. For years, Chrome provided server-side speech recognition for free through this API — Caption.Ninja just hooked into it.

That's changing. Google is migrating Chrome's speech recognition from the legacy server-side service to on-device SODA (Speech On-Device API) models. During the migration, the old service is being throttled or made unavailable, and the new on-device service isn't shipping reliably to all users. Multiple Chromium bug reports (issue 40286514, 40948113) document the breakage. Brave Browser has filed a similar issue — on-device SpeechRecognition silently hangs in "downloading" state with no SODA component ever installing.

Caption.Ninja's own documentation now recommends switching to Microsoft Edge as a workaround. But Edge has its own broken SpeechRecognition API after version 111 (reported on Microsoft Q&A). The architecture is the problem — relying on a browser's free speech recognition for a streaming tool means you depend on what browser vendors decide to ship.

Why Edge isn't a real fix

Edge is Chromium-based — it inherits all of Chrome's speech recognition issues. The bug reports above affect Edge too. And Edge has its own additional bugs (the SpeechRecognition API breakage after version 111). Caption.Ninja's "use Edge" suggestion buys you maybe a week before the same issues hit there too.

Firefox and Safari don't support free speech recognition at all — so they're not options either.

The architectural fix: server-side speech recognition

The actual fix isn't a different browser — it's a different architecture. Tools that rely on browser-native speech recognition will keep breaking because they depend on free APIs that Google is no longer prioritizing.

StreamTranslate uses Deepgram Nova-3 for speech recognition. That runs on dedicated servers (not in the browser) with a 99.9% uptime SLA. Audio streams from your mic to Deepgram, transcription comes back. The browser doesn't matter — Chrome, Safari, Firefox, Edge all work the same.

This costs more per minute than free Web Speech API ($0.0043/minute vs free), but it doesn't break. We charge for it because we pay for it.

StreamTranslate vs Caption.Ninja — the honest comparison

AspectStreamTranslateCaption.Ninja
Currently working in Chrome?✅ Yes (Deepgram is independent)⚠ Failing (Web Speech API being deprecated)
Works in Safari?✅ Yes❌ No (Safari has no Web Speech API)
Works in Firefox?✅ Yes❌ No (Firefox has no free Web Speech)
Free option6-hour trial, no cardFree forever (when working)
Languages out of the box30+ (incl. Japanese, Korean, Mandarin, Arabic, Hindi)17 European via Bergamot · BYOK for others
SetupOne URL paste in OBSConfigure URL parameters, optional API key
Cost (casual streamer)From $9.99 one-timeFree
Cost (asian languages)Included in any planBYOK Google Cloud (~$20/1M chars)
Speaker diarization🔜 Coming soon❌ Single mic only
Dual-language display✅ Pro+❌ Single only
Real-time word-by-word✅ Yes❌ Sentence-by-sentence
Hallucination filtering✅ YesLess needed (Bergamot is small MT)
PrivacyAudio sent to DeepgramAudio stays on your machine
Open sourceNoYes
Self-hostableNoYes

Caption.Ninja wins on cost and privacy. We win on reliability, browser support, language coverage (Japanese/Korean/Mandarin/Hindi/Arabic), and managed setup. Different tools for different streamers.

If Caption.Ninja is currently down — try us free

If you're hitting the Chrome speech recognition error right now and need something that works for your stream tonight: launch StreamTranslate — 6 hours free, no card. Setup is one URL pasted into OBS as a Browser Source. Same workflow as Caption.Ninja, different speech recognition backend.

Frequently asked questions

Will Caption.Ninja fix this?
It depends on Steve Seguin (the maintainer) migrating Caption.Ninja off the Web Speech API to a paid speech-to-text backend. That would mean the free tier wouldn't be free anymore. Until then, the issues will continue as Chrome's Web Speech API gets less reliable.
Is this affecting only Caption.Ninja?
No. Web Captioner (another popular free tool) shut down in October 2023 partly due to this. Any tool using browser-native Web Speech API is affected: Zip Captions, basic OBS browser-source caption scripts that rely on `SpeechRecognition`, etc.
Does StreamTranslate have outages?
Rarely. We use Deepgram with a 99.9% uptime SLA + multi-provider translation fallback (DeepL, Google, Azure). Our last major outage was < 4 hours total in 2026.
Can I still use Caption.Ninja for Bergamot translation while using StreamTranslate for speech?
Technically yes but redundant. StreamTranslate handles both speech recognition and translation in one URL. No need to chain tools.
Is there a Linux-friendly path?
StreamTranslate works on Linux via any browser that supports MediaRecorder (Chrome/Chromium/Firefox/Brave on Linux all work). Caption.Ninja struggles on Linux because Web Speech API is limited there.

Free 6-hour trial, no credit card

If Caption.Ninja is down for you right now, paste this in OBS and you're live in 60 seconds with translated captions.

Launch StreamTranslate →