Talk:OpenSpeaks/Tools/Subtitler

---

Generate stuck at 0%

Looks like a very useful tool. However, when I click on the Generate button for a file on Commons, it doesn't load (stuck at 0%). Prototyperspective (talk) 12:20, 6 April 2026 (UTC)

Hi @Prototyperspective, thanks for sharing your feedback. Noted and we'll update here as this is fixed. Psubhashish (talk) 17:17, 7 April 2026 (UTC)

Hi @Prototyperspective, @Ranjithsiji has recently fixed this and a few other issues. Please check and let us know. Psubhashish (talk) 06:48, 16 May 2026 (UTC)

Feedback

This is a great tool, I'm loving it! The subtitles editing process is quite intuitive thanks to the visual design that makes it easy to understand how time stamps work in the video timeline. The silence detection wasn't for me very useful (at least for my videos), so I decided to add segment by segment and make them fit with the timeline. As you get better at this, it gets easier and faster. Separately, I generated automated transcriptions with Whisper web, but I'm looking forward for something like this integrated in the tool soon. Finally, the integration with Commons is so smooth and fast! I have just a question: how can you mark a Draft project as Done in your Dashboard? Thank you! Mariana Fossatti (WK?) (talk) 15:58, 3 June 2026 (UTC)

Hi @Mariana Fossatti (WK?): Thank you much for using Subtitler and sharing these valuable input. Ranjith and I are discussing how to fix the silence detection issue. We have been discussing ASR for some time and might integrate Whisper as a starting point. Most Indigenous languages most likely will need their own finetunes, if existing, but won't be supported. Alternatively, we can try to figure out how to give the option to the user if they want to use Whisper or the current option, telling them the limitations of each one. We plan to have an in-tool doc page explaining these things in the next phase (July onwards). Psubhashish (talk) 13:36, 4 June 2026 (UTC)

Load from link

@Psubhashish is there a way i can link to it, with the video url in the link, that would make it load that video in the editor ? Would probably be wise to just prefill the "Search Commons" field or something, so that a user action is still required to load the file etc. With all these AI bots varying the URL you might otherwise be hit with a lot of scraping attempts for all sort of various urls. —TheDJ (talk • contribs) 07:43, 11 August 2026 (UTC)

Hi @TheDJ, just the Commons part seems to be possible and it might not warrant scraping. Note that we don't have a server for this tool. Everything happens on the browser, with some cache stored in the browser just to resume a session. So, technically, a scraping attempt would be pointless. Psubhashish (talk) 04:10, 15 August 2026 (UTC)

@Psubhashish "So, technically, a scraping attempt would be pointless." well... AIs don't care about the point, they simply throw a wide net. If the url can vary and the 'browser' will just start loading wikimedia resources automatically, there will still be impact on the foundation resources. —TheDJ (talk • contribs) 13:09, 26 August 2026 (UTC)

To illustrate, we've removed some 'expensive' links toward video files from the anonymous version of the pages recently. Scraping tools would simply follow them and download multi terabyte files. By removing those links and hiding them a bit, load for video downloads decreased quite a bit. —TheDJ (talk • contribs) 13:11, 26 August 2026 (UTC)

Thank you for sharing that example. It's something to deal with caution. Noting. We haven't deployed anything yet but it's helpful to know what you've shared kindly. Psubhashish (talk) 13:32, 26 August 2026 (UTC)