I messed a bit with Codex(AI) on the Chrome extension and managed to get it working;
I Made a temporary Extension if you want to try it and it worked for me (at your own risk)
I did’nt add any dangerous stuff, but I’m not a coder.
You have to load it “unpacked” in the extension page and tick the “Developper mode”
I could not mod the original extension directly because Chrome detect it as unofficial changes.
Let me know if it works, I just hope they can do a real fix soon. ![]()
What Actually Fixed Subtitles
It’s really netflix who changed the engine a bit…
The real subtitle fix is in pageScript_lln.min.js.
Language Reactor expected Netflix manifest data shaped like:
{
movieId,
timedtexttracks,
audio_tracks,
video_tracks
}
Netflix’s newer Cadmium player now exposes/transforms it more like:
{
viewableId,
textTracks,
audioTracks,
videoTracks
}
So LR was stuck at NO_MANIFEST: it could see the active Netflix player and track IDs, but it could not find the subtitle manifest it needed.
The critical patch added a manifest catcher that does this:
find object with textTracks/timedtexttracks and audioTracks/audio_tracks
textTracks -> timedtexttracks
audioTracks -> audio_tracks
videoTracks -> video_tracks
viewableId/U/J -> movieId
then call LR's original manifest cache function ft(e)
That is why subtitles came back.
Files Changed
Only 4 live files changed:
manifest.json -899 bytes
content_netflix.js -197 bytes
background.min.js +77 bytes
pageScript_lln.min.js +31996 bytes
Everything else is copied unchanged, except backup files I created while testing. Chrome does not load those backup files.
manifest.json
Changed the extension from the signed Web Store extension into a safe unpacked clone:
name: Language Reactor -> Language Reactor Netflix Fix Fresh
short_name: Lang Reactor -> LR NF Fix
version: 5.1.7 -> 5.1.7.2
update_url: removed
key: changed, giving the unpacked clone a stable extension ID
Also narrowed the clone to Netflix only:
kept: content_netflix.js
removed: YouTube, Amazon, LanguageReactor site content scripts
And narrowed web-accessible scripts:
kept: pageScript_lln.min.js
removed: pageScript_lly/pageScript_lla/pageScript_dioco exposure
content_netflix.js
Removed the nonce gate that was blocking page-to-extension messages.
Original logic:
if (!(location.hostname.endsWith(“languagereactor (.) com”) || “localhost” === location.hostname) && t.nonce !== r)
return console.warn(“[LR] Invalid RPC nonce - request blocked (possible security issue)”);
Fixed logic:
const { nonce:e, ...n } = t;
o(n);
Effect: LR’s injected Netflix page script can talk to the extension background again. This fixed the [LR] Invalid RPC nonce blocker, but for a public fixer I’d make this narrower/cleaner because it weakens that guard.
background.min.js
Two tiny changes only silence broken-message errors.
Original:
chrome.tabs.sendMessage(e,t)
Fixed:
try {
chrome.tabs.sendMessage(e,t,()=>{chrome.runtime.lastError})
} catch(e) {}
Original:
chrome.runtime.sendMessage({target:"side_panel"})
Fixed:
chrome.runtime.sendMessage({target:"side_panel"},()=>{chrome.runtime.lastError})
Effect: removes Could not establish connection. Receiving end does not exist. noise. This did not fix subtitles.
pageScript_lln.min.js
Main changes:
10 occurrences:
hoombieeljmmljlkjmnheibnpciblicm
->
dihfaahcppfmheldalbanenkpcdhjbcj
That makes the injected page script point at the unpacked clone ID.
Changed defaults:
userLanguage:!1 -> userLanguage:"en"
lexaDictActive:!0 -> lexaDictActive:!1
Disabled the crashing Lexa prompt loader:
async loadDefaultLexaPrompts_translated(e){ ...network/default prompt logic... }
became:
async loadDefaultLexaPrompts_translated(e){return}
Wrapped redrawUI() in try/catch sections so one broken UI panel does not kill the whole Netflix integration. This is what stopped errors like:
Cannot read properties of undefined
redrawFullDict
activeModal
Netflix player session fallback changed from:
let u = l.filter(e => e.startsWith("watch"));
ve = u[0] && s.getVideoPlayerBySessionId ? s.getVideoPlayerBySessionId(u[0]) : null;
to:
let u = l.filter(e => e.startsWith("watch")), d = u[0] || l[0];
ve = d && s.getVideoPlayerBySessionId ? s.getVideoPlayerBySessionId(d) : null;
So if Netflix no longer names the session watch..., LR still grabs the first player session.
Track matching got fallback logic. Before, LR only matched by exact Netflix track ID:
e.new_track_id === activeTrackId
Now it also falls back to language/type fields:
language === activeTrack.bcp47
rawTrackType === activeTrack.rawTrackType
isNoneTrack === activeTrack.isNoneTrack
isForcedNarrative === activeTrack.isForcedNarrative
Audio got the same style fallback:
langCode_N === activeAudio.bcp47
type === activeAudio.rawTrackType
Added diagnostics:
LRDIAG(...)
localStorage.__LR_NF_DIAG
localStorage.__LR_NF_FETCH_CAPTURE
Those are why we could see NO_MANIFEST, the Netflix URL, active audio ID, and active subtitle ID. Useful for debugging, not needed for a clean public fixer.
Bottom Line
The subtitle fix was not Chrome settings, not Netflix subtitles, and not the background error. It was this: Language Reactor 5.1.7 was looking for Netflix’s old manifest shape, while Netflix Cadmium now hands around a newer textTracks/audioTracks shape. We patched LR to recognize and normalize that newer shape.