Bug 499012 - Kdenlive flatpak downloads whisper model to wrong location
Summary: Kdenlive flatpak downloads whisper model to wrong location
Status: CONFIRMED
Alias: None
Product: kdenlive
Classification: Applications
Component: Title Clips & Subtitles (show other bugs)
Version: 24.12.1
Platform: Flatpak Linux
: HI normal
Target Milestone: ---
Assignee: Jean-Baptiste Mardelle
URL:
Keywords: triaged
Depends on:
Blocks:
 
Reported: 2025-01-22 15:26 UTC by frafortunelli
Modified: 2025-03-07 16:48 UTC (History)
3 users (show)

See Also:
Latest Commit:
Version Fixed In:
Sentry Crash Report:


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description frafortunelli 2025-01-22 15:26:28 UTC
SUMMARY
Speech to text setup on kdenlive flatpak downloads whisper model to wrong location, thus speech to text does not work without manual copying of file. 

STEPS TO REPRODUCE
1. settings -> configure kdenlive -> speech to text -> install multimodal translation 
2. install whisper turbo model 
3. run speech to text 
4. whisper model is downloaded to ~/.cache/whisper but kdenlive cannot find it, and searches for it in ~/.var/app/org.kde.kdenlive/cache/whisper

OBSERVED RESULT
Speech to text does not work

EXPECTED RESULT
Speech to text works 

SOFTWARE/OS VERSIONS
Operating System: Bazzite 41
KDE Plasma Version: 6.2.5
KDE Frameworks Version: 6.10.0
Qt Version: 6.8.1
Kernel Version: 6.12.9-203.bazzite.fc41.x86_64 (64-bit)
Graphics Platform: Wayland

ADDITIONAL INFORMATION
Copying the whisper file from ~/.cache/whisper to /.var/app/org.kde.kdenlive/cache/whisper, speech to text now works.
Comment 1 emohr 2025-01-22 18:36:59 UTC
Thank you for this detailed report.

@JBM another path issue for Wisper on flatpack.
Comment 2 Bruno Cabral 2025-01-29 03:34:57 UTC
Just had the same issue.
Wanted to add that if one creates folder ~/.var/app/org.kde.kdenlive/cache/whisper in advance, the model is downloaded in the correct location.
Comment 3 Jean-Baptiste Mardelle 2025-01-30 09:57:12 UTC
Git commit e176514b6733a9b33fb7180f54ff89ddf344e6a8 by Jean-Baptiste Mardelle.
Committed on 30/01/2025 at 09:56.
Pushed by mardelle into branch 'master'.

Try to fix Whisper models folder on Flatpak

M  +13   -3    src/pythoninterfaces/speechtotextwhisper.cpp

https://invent.kde.org/multimedia/kdenlive/-/commit/e176514b6733a9b33fb7180f54ff89ddf344e6a8
Comment 4 Jean-Baptiste Mardelle 2025-01-30 09:57:38 UTC
Git commit 1d468e892e8fd034d020f1e4c7c1144b573bbcf0 by Jean-Baptiste Mardelle.
Committed on 30/01/2025 at 09:49.
Pushed by mardelle into branch 'release/24.12'.

Try to fix Whisper models folder on Flatpak

M  +16   -3    src/pythoninterfaces/speechtotextwhisper.cpp

https://invent.kde.org/multimedia/kdenlive/-/commit/1d468e892e8fd034d020f1e4c7c1144b573bbcf0
Comment 5 Awadea 2025-03-07 16:48:29 UTC
With the Appimage on CachyOS (Arch based) no speech model installed and the download won't start because of no existing folder under .cache/whisper. I've created this folder and the download starts correctly and kdenlive finds the files after downloading.