Case 1 – Incomplete subtitle generation (single-language output) I used Ray Local to generate English subtitles for a 45-minute documentary containing both Danish and English audio. The result had large gaps: whole sentences, sometimes 30–40 seconds of continuous speech, were simply missing. The gaps occurred mainly in the Danish portions — particularly right after the dialogue switched from English to Danish. English-language sections were transcribed correctly.
Case 2 – Broken translated subtitles (multi-language output) I then tried to generate both German and English subtitles from the same source. The German output was essentially unusable — garbled text that no German speaker could meaningfully follow. It’s unclear to me whether Ray translated the already-generated English subtitles or re-transcribed the original audio; either way, the result was comprehensible in neither.
Expected behaviour: Complete transcription of all spoken content regardless of language switches, and coherent translations.
Question: Are language switches mid-recording a known limitation? And is there a recommended workflow for generating translated subs from a mixed-language source — e.g., first creating a clean transcript, then translating it?
I don’t know what Ray Cloud used - if there was a choice I surely used Sunlight for better results. I don’t find any content ID in the Logs, but I can tell you when it took place in Ray Cloud (CET):
Aufgabe: Untertitel
Sprachen: EN-US
Ort: Ray Cloud
Ergebnis: Fertig
Gestartet: 23.10.26 13:24
Beendet: 23.10.26 14:51
Dauer: 1:27:12
I don’t want people who develop software remote connect to my computer … I actually just expect software to work as sold/promised.
And I find it kind of scary that you publish a software with a built in remote connect in it.
“Allow support sessions” is enabled by default. Opt-out is bad design: a channel through which a server can trigger tests, change settings, and initiate downloads should be opt-in.
This actually makes me want to uninstall this app asap …