Community Wishlist/Wishes/A tool for auto-transcription to speed up the creation of TimedTexts subtitles for videos on Commons/en
Description
On Wikimedia Commons, users do video transcriptions all by hand which takes a lot of time. At the same time on YouTube one can just generate them and then only needs to modify the transcription (check if there's any errors) or alternatively provide the transcription text and let YouTube set the timings (which sentences are said when). Contributors would be far more productive if they could benefit from auto-transcription tools, time they could use for other contributions to Wikimedia projects or for contributing more transcriptions.
The tool could also be used to display auto-captions which can be very useful if otherwise there are none for a video, especially if the video has bad audio-quality or is in a language the user doesn't speak well.
Even if there are multiple errors in the transcription, that would be very useful and time-saving as it would at least have all those timings and manually correcting machine transcribed subtitles is much quicker and easier than writing it all from scratch. I think for English at least auto-transcription usually works quite well if the pronunciation and audio-quality in the video is fairly okay.
This has been asked multiple times on Wikimedia Commons, a few links and some info here.
There are probably two ways this could be implemented:
- Auto add automatic transcriptions in the language of the video audio to files and display that in a way so that users know it's machine transcribed and may contain errors (like YouTube has it too) – e.g. "Italian (auto)" and have these subtitles disabled by default (and I think they should always be disabled by default but especially these)
- As only a tool for contributors to use when they create transcriptions when going to a video's TimedText page – e.g. at the top of the page there could be a button saying "Generate transcription" with info that the user should check them and a button "Add transcription without timings" where it only sets the timings like
00:00:41,000 --> 00:00:47,050
Obviously, if method 1 is implemented, that should only be done for languages where machine transcription works sufficiently well and if there are concerns about errors, an extra warning could be displayed at start or when clicking that subtitles button saying e.g. "These captions have been generated and may be faulty."
This should be a well-integrated tool enabled by default (once tested and functional) or part of the software (and the latter would probably be better). It shouldn't be some gadget that has to be enabled manually by registered users that nobody knows exists.
This is a requested feature that hasn't been included in the WMC's recent technical needs survey as it isn't as urgent as the proposals made there. It would just be very useful and save contributors a lot of time while improving multilingualism/internationalization.
Update: by now a found a tool that is both open source and mostly highly accurate that could be used. It can also be used to redub videos but this is more difficult than transcripts and then translating these. You can install SoniTranslate locally as described here and here. Some changes could be made so one can create subtitles for many languages at once or so that there is a maximum text length per segment, minimum space between segments, and so on. I suggest that this tool is used/integrated in some way, for example a button "add machine transcription" that adds many transcriptions at once but displays (auto) next to the language in the closed captions until it has been checked and if needed adjusted. I did not add closed captions to the example redubbed video on the right and other ones where I have the transcriptions because they should not be displayed by default, annoying most viewers; see phab:T371727.
Related wish: Add a subtitling tool for easy subtitling and translating
Assigned focus area
Unassigned.
Type of wish
Feature request
Related projects
Wikimedia Commons, Wikipedia
Affected users
Wikipedia readers who look for subtitles especially people not understanding the video's language well; and WMC contributors who seek to add timedtexts or save time doing so to e.g. add more of them
Other details
- Created: 11:59, 26 July 2024 (UTC)
- Last updated: 16:40, 9 August 2024 (UTC)
- Author: Prototyperspective (talk)