[GH-ISSUE #4997] [BUG]: Unsupported model type: whisper #3136

Closed
opened 2026-02-22 18:32:46 -05:00 by yindo · 3 comments
Owner

Originally created by @Persimonk on GitHub (Feb 12, 2026).
Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4997

How are you running AnythingLLM?

AnythingLLM desktop app

What happened?

I know that there have been at least 2 reports of this bug already, but I have seen them and I can't find a solution. I have an AMD processor but I am running the desktop app, not Docker.

Whenever I try to use the STT, it says "Unsupported model type: whisper". I tried downloading the model separately, but it makes no difference. However, what I did find out, is that after switching the language (and forcing the download and usage of the multilingual tiny whisper), the error disappears. Therefore I am thinking that there is some problem specifically within the whisper-tiny.en model.

Are there known steps to reproduce?

Use the "Speak your prompt" button while language is set to English.

Originally created by @Persimonk on GitHub (Feb 12, 2026). Original GitHub issue: https://github.com/Mintplex-Labs/anything-llm/issues/4997 ### How are you running AnythingLLM? AnythingLLM desktop app ### What happened? I know that there have been at least 2 reports of this bug already, but I have seen them and I can't find a solution. I have an AMD processor but I am running the desktop app, not Docker. Whenever I try to use the STT, it says "Unsupported model type: whisper". I tried downloading the model separately, but it makes no difference. However, what I did find out, is that after switching the language (and forcing the download and usage of the multilingual tiny whisper), the error disappears. Therefore I am thinking that there is some problem specifically within the whisper-tiny.en model. ### Are there known steps to reproduce? Use the "Speak your prompt" button while language is set to English.
yindo added the possible bug label 2026-02-22 18:32:46 -05:00
yindo closed this issue 2026-02-22 18:32:46 -05:00
Author
Owner

@timothycarambat commented on GitHub (Feb 13, 2026):

Whenever I try to use the STT, it says "Unsupported model type: whisper". I tried downloading the model separately, but it makes no difference. However, what I did find out, is that after switching the language (and forcing the download and usage of the multilingual tiny whisper), the error disappears. Therefore I am thinking that there is some problem specifically within the whisper-tiny.en model.

This is odd, i have cannot get this replication to occur even when swapping languages and such. The model from HF is pulled into the application storage (in the window for electron) which is why downloading it does not impact this. It would appear that whisper-tiny.en might have failed to download some part of its config/file and is crashing due to that missing file in storage.

For this, we have a new control on the settings page to clear the storage of that so you can freshly re-download the file in the next release. Depending on your connection at the time it could have not pulled fully and is failing to load or is corrupt. I know HF has blocks for people on VPNs or if you are based in China - specifically we see that a lot and have to mirror some critical path assets on our CDN to patch that.

@timothycarambat commented on GitHub (Feb 13, 2026): > Whenever I try to use the STT, it says "Unsupported model type: whisper". I tried downloading the model separately, but it makes no difference. However, what I did find out, is that after switching the language (and forcing the download and usage of the multilingual tiny whisper), the error disappears. Therefore I am thinking that there is some problem specifically within the whisper-tiny.en model. This is odd, i have cannot get this replication to occur even when swapping languages and such. The model from HF is pulled into the application storage (in the `window` for electron) which is why downloading it does not impact this. It would appear that whisper-tiny.en might have failed to download some part of its config/file and is crashing due to that missing file in storage. For this, we have a new control on the settings page to clear the storage of that so you can freshly re-download the file in the next release. Depending on your connection at the time it could have not pulled fully and is failing to load or is corrupt. I know HF has blocks for people on VPNs or if you are based in China - specifically we see that a lot and have to mirror some critical path assets on our CDN to patch that.
Author
Owner

@Persimonk commented on GitHub (Feb 13, 2026):

You are probably right, I did a reinstall for another reason and it seems to be okay. However a setting that would allow to redownload files after suspecting corruption would be amazin.

@Persimonk commented on GitHub (Feb 13, 2026): You are probably right, I did a reinstall for another reason and it seems to be okay. However a setting that would allow to redownload files after suspecting corruption would be amazin.
Author
Owner

@timothycarambat commented on GitHub (Feb 13, 2026):

That is exactly the solution! Will be in next patch so we can wipe it from storage and then on next execution it will re-pull (and hopefully function)

@timothycarambat commented on GitHub (Feb 13, 2026): That is exactly the solution! Will be in next patch so we can wipe it from storage and then on next execution it will re-pull (and hopefully function)
yindo changed title from [BUG]: Unsupported model type: whisper to [GH-ISSUE #4997] [BUG]: Unsupported model type: whisper 2026-06-05 14:50:31 -04:00
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Mintplex-Labs/anything-llm#3136