Community Wishlist/Wishes/Fix the main problems (speech slowdown) in the open source SoniTranslate used to redub videos

Fix the main problems (speech slowdown) in the open source SoniTranslate used to redub videos Open

Description

Here at 0:34 and 1:05

Here at 0:47

I know this isn't a Wikimedia tool. But if it's open source and used for Wikimedia purposes, then it effectively is a tool used by Wikimedians for Wikimedia, and some of the problems within it affect these. There is no need to develop a new tool from scratch if an existing freely licensed one already exists. If anything, it could be forked if necessary to be more Wikimedia-specific and get some things fixed.

SoniTranslate is written in Python, free software under the Apache-2.0 license, and may be the best open-source redubbing/transcription/AI-voice tool.

I've used it extensively to create podcast audios of Wikipedia articles that sound very good and are usually more listenable than audios spoken by amateur narrators, and I've partly turned to listening to these instead of podcasts.

The tool can also be used to transcribe videos (add subtitles) and to redub videos into other languages, which saves lots of time compared to doing it fully manually and usually works quite flawlessly.

However, there are some issues with it. They seems like minor issues but they're quite annoying and kind of screw up the quality. The biggest one is that sometimes the tool slows down the speech. Please see the two example videos on the side, which I redubbed from German to English using the tool – at the specified timestamps the speech slows down.

Fixing this and improving this tool to be more suitable for Wikimedia redubbing could greatly benefit Wikimedia audiovisual content & multilingualism. The Commons category Videos by Terra X alone contains so many high-quality educational videos that could be redubbed. Whatever efforts have been made in Wikimedia to add subtitles and translate videos so far, the use of this tool could be used to achieve double that in just one year.

Here is the issue: Speed of the translated audio is slowed down. (#89)

I've listed further major issues at c:Help:AI video dubbing#Known problems.

Assigned focus area

Unassigned.

Type of wish

Bug report

Related projects

Wikipedia, Wikimedia Commons, Wikidata

Affected users

Wikipedia readers, Commons users

Other details

Created: 14:11, 14 April 2025 (UTC)
Last updated: 13:08, 17 April 2025 (UTC)
Author: Prototyperspective (talk)