CaptionGo v1.1 Release Notes

CaptionGo v1.1 makes Live Captions work on significantly more video sites and streaming platforms. The speech-to-text engine has been completely rebuilt to detect audio sources more reliably, handle encrypted streams, save bandwidth by downloading only audio data, and render subtitles more cleanly on complex web pages.

• Live Captions status for CaptionGo Cloud: The status indicator in the top-right corner now shows caption progress with the built-in CaptionGo Cloud too. Click it to check today's remaining free trial time, switch the recognition language, or stop Live Captions
• Pre-generated subtitles on more sites: Live Captions can now generate subtitles ahead of playback on more video sites, and they stay in sync with the speech even when you turn it on midway through a video
• Advanced Deepgram settings: Fine-tune recognition in the Deepgram provider settings — smart formatting, profanity filter, and key terms that help recognize names and technical words more accurately, plus separate sentence-break options for real-time and pre-generated captions

• Easy-to-read captions for long speech: Long stretches of speech are now split at sentence ends and natural pauses into short captions, instead of one long block without breaks. Sentences are also more complete, with fewer missing endings (you can turn off "Split Long Captions" in settings)
• More accurate automatic language detection: Music intros, silence and other parts without speech no longer throw off detection, and a language is only used once the result is reliable
• Step-by-step language confirmation: When generating subtitles ahead of playback, Live Captions now recognizes one segment at a time and confirms the spoken language before speeding up. The progress count only goes up once subtitles are actually ready, and processing pauses immediately if the language looks wrong — saving your quota
• Same experience with CaptionGo Cloud and your own Deepgram key: Recognition language, capture mode, silence filter and other settings now apply to the built-in CaptionGo Cloud too, and subtitles are prepared just as far ahead
• Clearer language detection messages: Each step — detecting, confirming, correcting or failing — shows a short message with the language name instead of a code, and you can pick a language right away
• Error details in the Live Captions menu: When Live Captions can't start or stops, the menu now shows the reason instead of a generic error
• More accurate real-time caption timing: Subtitles no longer go missing or out of sync after a brief network drop
• Cleaner bilingual subtitles: Earlier lines no longer reappear and push the latest line off screen

• Fixed the beginning of long sentences being cut off in real-time captions
• Fixed the "language switched" message disappearing right after it appeared
• Fixed progress getting stuck on the same number when the speech recognition service can't be reached, followed by an unnecessary switch to real-time mode — you now get a "check your network" message within about 40 seconds, with one-click Retry
• Fixed no message appearing when a real-time connection fails
• Fixed audio still being uploaded in the background after Live Captions was stopped, which used up quota
• Fixed Live Captions not working after signing in until the page was refreshed

• Smarter language detection: When Live Captions can't reliably detect the spoken language, it now pauses and prompts you to select a language manually instead of guessing — saving your recognition quota
• Better Chinese/Japanese/Korean recognition: Improved accuracy when distinguishing between Chinese, Japanese, and Korean audio
• Language re-detection on seek: Seeking to a different part of the video now restarts language detection from the new position — ideal for skipping music intros or navigating to speech sections
• More reliable audio processing: Improved compatibility with a wider range of video formats, reducing cases where audio download would get stuck at "Downloading audio…" or fail to produce subtitles
• Audio download progress: Now shows real-time download progress with speed (e.g. "Downloading audio… 35% (120 KB/s)") so you know exactly what's happening
• Better reliability on slow networks: Audio download now automatically retries with optimized settings on slow or unstable connections
• Processing progress between segments: Shows progress indicators (e.g. "Processing audio… 3/7") between subtitle segments, only when no subtitle is currently displayed
• AI translation quality upgrade: Refined translation prompts for better subtitle coherence, larger context window (up to 30 previous lines), and video metadata (title, synopsis) now informs the translation for more accurate results
• ASR-aware AI translation: AI translator now adapts its approach when translating speech recognition subtitles, producing more fluent results for imperfect source text

• AI translation privacy control: New toggle in Translation Settings to control whether video context (title, URL, description) is sent to AI translation — enabled by default, can be turned off for privacy

• Fixed video black screen when opening video file URLs directly in the browser (e.g. .mp4 links)
• Fixed MP4 audio extraction downloading nearly the entire file for certain video formats (e.g. lecture/educational videos), now intelligently extracts only audio data
• Fixed subtitles stopping midway through longer videos when the predownload window was fully processed
• Fixed "Preparing subtitles…" message stuck on screen after seeking to an already-processed section
• Fixed predownload progress indicator getting stuck and not reaching 100%
• Fixed language selection badge not appearing when manual language selection was needed
• Fixed translation inaccuracies caused by repeated phrases across different subtitle segments sharing cached translations
• Fixed membership status not refreshing after browser restart

• Fixed Live Captions predownload restarting repeatedly on sites using Wistia player (e.g. online course platforms)

• Better predownload stability — fewer unnecessary restarts, avoids re-processing already-recognized audio segments

• Live Captions on more sites: completely rebuilt speech-to-text engine that automatically detects audio sources from HLS, DASH, and MP4 video streams
• Automatic video stream discovery: detects video manifests and audio URLs by monitoring page network traffic, so Live Captions works even on sites without explicit audio source configuration
• Encrypted video support: Live Captions now works on sites using AES-128 encrypted audio, including more streaming platforms with protected content
• Bandwidth-efficient audio download: extracts and downloads only the audio track from MP4 files instead of the full video, reducing data usage significantly
• TikTok and hotlink-protected sites: automatically adds the correct request headers so audio predownload no longer fails on sites like TikTok that block external requests

• Live Captions no longer shows "downloading…" or "processing…" messages once subtitles have started appearing — only errors are shown
• Improved subtitle style isolation using Shadow DOM, preventing conflicts with complex web page styles
• Audio predownload settings can be updated remotely without requiring an extension update
• ASR language detection prompt: when Live Captions returns consecutive empty results, a language detection banner appears with an inline language picker for quick switching
• Automatically falls back to realtime mode when predownload encounters DRM-protected content or fails to obtain an audio source
• AI translation now retries automatically to reduce missed translations caused by model initialization race conditions

• Fixed an issue where dragging subtitles to the top of the video caused the drag handle to overflow, making the subtitles impossible to reposition
• Fixed caption cache mismatch on pages with multiple videos (e.g. Twitter/X) where subtitles could appear on the wrong video