As per the title: Kdenlive has an option in the “speech to text” section with an option to download a speech model. Turbo is advised.
Anyway, mine checks for available models and then starts to download.
Unfortunately, it doesn’t. I let it sit overnight but to no avail.
The progess bar immediately disappears and nothing happens.
So, I intend to download it seperately and put it where it belongs.
So I put the ‘large-v3-turbo.pt’ in …/bin but Kdenlive did not see it.
Unpacked the .pt file. Still no luck.
Updated Whisper via config in Kdenlive and now I have a folder in ~/.local/share/kdenlive/venv/Lib called large-v3-turbo.
Somehow Kdenlive still asks me to download a speech model.
Which does not work.
Well, I can still use the AppImage which does work but I’d rather use the flatpak.
In addition and for making things visual, here is a screenshot.
It stays that way even overnight.
I used to use the “Veronica explains” way before but that doesn’t work anymore. So I manually deleted all files and folders before reïnstalling Kdenlive 24.12.2.
Are you using flatpak? That’s explain it can’t see it as flatpak runs the apps in sandbox. You should get your system package of Kdenlive instead.
Or try mess flatseal to set the permission of Kdenlive.
Yes, kinda stupid I didn’t mention it before. I’ve been using Flatpak for a long time.
The Flatpak version has an install option in Configure>Speech to text since the latest update. Before that I used the method described in the documentation. The, what I call, Veronica explains method.
Always worked flawless but not anymore.
The version in the repo is always several behind. Hence the Flatpak.
Sorry for not being clear from the start. I’ll try to change the title to be clear.
I rarely use Flatpak as I’m using rolling distro, so I don’t have to ran into such problems .But you need to use Flatseal to set the right permisssion of Kdenlive first before it can see outside its box. After that you need to set up custom path in kdenlive where to look for the plugins.
That is good for you. I use Mint and have used it for over ten years.
Nevertheless, thank you for at least making the effort to try and help me.
By the way, Whisper is correctly installed from within the Flatpak Kdenlive.
So only the speech/language models won’t download/install.
It is my belief that the latest Flatpak should have the entire proces ‘automated’ when required. At least that is how I understand.
It uses a Python venv to accomplish things.
Okay, so the solution was very easy.
Reading the releasenotes for 24.12.2 I saw there was a problem with the folders for the speech models.
Somehow I must have not deleted all existing folders before re-installing the new Flatpak. I.E. I forgot a few.
If anyone ever runs into the same problem as I did:
create a folder named whisper in ~/.var/app/org.kde.kdenlive.org/cache.
That’s all. Now restart Kdenlive and download the speechmodel.
It will now download and work perfect.
This bug was reported before so I could use the bugreport to solve the issue.
Thanks to everyone who thought with me and to Storm for replying.