Transcript doesn't render
Quite frequently (easily a couple of times a day) I’ll be trying to write something and all that will be captured is the first word. Very frequently this gets rendered as something like - Perspective or - Tests or something like that (with the hyphen being something that is very often present). I’m not sure why this is occurring but it happens with a degree of frequency that means I have worry about whether what I’ve just said will actually be captured and rendered.
Macbook M5, 128GB - Small whisper model, and then mistral running locally as custom AI pathway.
Log in to comment and vote
Comments7
ajfisher
May 7
So after the update that dropped a couple of days ago it definitely improved but I’ve had it happen a couple of times over the last few days. Listening back to the audio, it all sounds totally fine - I can share this one as it doesn’t have anything sensitive in it, unlike the other couple of times.
It recorded fine, did the little “processing” step and then resulted in a “ “ entered in the textfield I was prompting into. I have turned off any AI post processing summarisation / clean up as well so as to remove moving parts.
What’s interesting, is if I take this and upload it to the transcribe tool in vowen it renders the text almost instantly, whilst when I did it “live” it was hanging on the “processing” animation for a little while (10-15s for a pretty short snippet) before rendering anything at all. I feel like that’s relatively common but I couldn’t say that definitively as I just noticed that on this try by doing the comparison.
vowen_2026-05-07_11-04-03.wav
693 KB• Audio
Vowen
May 7
I see, we’ll investigate it once again and get back to you, I’ve reopened this issue.
We have a resource efficient mode if you are using local models which can be found in “Settings > General” at the bottom, you can try switching on the setting and let us know if that helps. What we suspect is happening is that the local model server might be crashing midway and causing this, we are adding more stronger guardrails to prevent this from happening. So turning on “resource efficient mode” should be almost instant as you’d reported when you used the “transcribe tool” with the app.
ajfisher
May 9
Okay - will give this a crack over next several days and give you a sense of impact.
Vowen
May 6
We are closing this, due to inactivity. Please reopen this if you are still facing this issue.
ajfisher
May 7
Not sure how to reopen but have posted an example that happened this morning.
Vowen
May 1
@ajfisher We will look into this issue and get back to you.
For now, can you temporarily turn on “Save audio recordings” setting from “Settings > Recording” section of the app? And when do notice this happening, can you let us know if the audio recording file has the complete audio you’ve recorded, or does it get cut off? The recordings should be in “~/Documents/Vowen Recordings” folder. This will help us understand if it’s related to the microphone or the underlying local model.
We will also switch to Small Whisper model and test things out and try to reproduce this from our end in the meanwhile.
Vowen
May 5
@ajfisher Let us know if you are still facing this issue