Why my live captions stopped showing up (a debugging checklist)

You clicked Start, picked a tab, and… nothing. The status bar shows “Listening” but captions never arrive. Or they show up for 5 seconds and freeze. Or they appear once and stop.

This is almost always one of 7 specific issues. Here’s the checklist to walk through, ordered by frequency.

By far the most common cause. Chrome’s tab/screen picker has a small checkbox at the bottom labeled “Share tab audio”. If you don’t tick it, only the video is shared — no audio. The captioner gets a stream but no sound, so no captions.

Fix: Stop, click Start again, and look for the checkbox. It’s unticked by default. On macOS the checkbox might say “Share audio”; on Windows it’s “Share tab audio.” Always tick it.

2. The audio source is silent or paused

LiveCaptionIt and similar tools have a silence guard that resets the buffer after 2.5 seconds of detected silence (RMS below a threshold). If the YouTube video is paused, the podcast hasn’t started, or the meeting is in a “waiting room” state, no audio = no captions.

Fix: Confirm audio is actually playing. Look for the volume indicator in your tab’s title bar (the small speaker icon Chrome shows on tabs producing audio). If it’s not there, your source isn’t producing audio yet.

3. The model is still downloading

On first use, the Whisper model downloads from the Hugging Face CDN. For Base it’s ~75 MB; for Small it’s 244 MB; for Large turbo it’s ~537 MB. On a slow connection this can take 30 seconds to several minutes. During this time the status bar shows download progress, not captions.

Fix: Wait for the status to change from “Loading model” to “Listening.” If it’s been over 3 minutes on Base, check your network tab in DevTools — the model fetch may be failing silently (CORS or network errors are reported separately).

4. WebGPU isn’t enabled

If your browser doesn’t support WebGPU (or you’ve disabled it), LiveCaptionIt falls back to WASM. WASM is functional but ~3-5× slower than WebGPU. On weak hardware the latency budget runs out before each tick completes, and captions appear to stall while the worker backlogs.

Fix:

Check chrome://gpu in Chrome — look for “WebGPU: Hardware accelerated”
On Linux, WebGPU may need a flag: chrome://flags#enable-unsafe-webgpu
On older laptops without dedicated GPU, switch to the Tiny (39 MB) model tier — it runs fine on WASM

5. The Whisper worker silently crashed

Web Workers can crash and the parent page sometimes doesn’t notice (the worker stops responding but doesn’t emit an error event by default). LiveCaptionIt has explicit error and unhandledrejection listeners inside the worker, plus a 120-second timeout on pipeline init, but custom code or browser quirks can still create silent hangs.

Fix: Stop, refresh the page, start again. If it consistently hangs, open DevTools → Console and look for messages prefixed [LiveCaptionIt: — the error monitor buffers recent errors there. Type __lcDebugErrors() to dump them.

6. macOS can’t capture system audio

This is a platform limitation, not a bug. Apple’s sandbox doesn’t let browsers capture system audio outside a specific tab. So picking “Entire screen” on macOS gives you video but no audio.

Fix:

Pick a specific Chrome tab instead, not “Entire screen”
For native macOS apps like Spotify desktop or Zoom desktop, you can’t capture them through the browser. Install BlackHole (free virtual audio device) and route the app’s audio through it into a Chrome tab capture
Or just use the web version of the app

7. The tab is producing protected audio

Some media (Netflix, certain Spotify content, some DRM-protected videos) is flagged as “protected” by the browser. The browser refuses to share its audio via tab capture even with consent. The tab capture silently returns a video-only stream.

Fix: Use a non-DRM source. YouTube, Vimeo, most podcasts, all meeting tools — these aren’t protected. Netflix and some music streams are. There’s no workaround; this is intentional DRM enforcement.

When none of the above apply

If you’ve checked all 7 and still no captions, the bug is likely in LiveCaptionIt itself. Open an issue with:

Browser + version (Chrome 130, Edge 130, etc.)
OS + version
What you were trying to caption (tab, mic, sample)
Output of __lcDebugErrors() from DevTools console
Any console errors with [LiveCaptionIt: prefix

We try to get back within a couple days. The captioning surface area is small enough that most reported bugs are reproducible from a short description.

One more debug tip

If you suspect the audio capture is the problem (not Whisper), open DevTools → Console and run:

const s = await navigator.mediaDevices.getDisplayMedia({
  video: true,
  audio: true,
});
console.log("audio tracks:", s.getAudioTracks().length);
s.getTracks().forEach(t => t.stop());

Pick your tab + tick “Share tab audio” in the picker. If audio tracks: 0, the picker didn’t share audio. If audio tracks: 1, capture is working — bug is downstream.

1. You didn’t tick “Share tab audio” in the picker