Whisper Tiny on Intel Mac Unusably Slow #575
-
|
I have whisper.cpp running locally with GPU disabled and it runs in close to real time. In FluidVoice, I have downloaded and activated the Whisper Tiny model and a simple "this is a test" voice prompt takes around 10 seconds to produce text. I have enabled the audio capture in the history settings, but I see no other way to debug this issue. The FluidVoice process memory foot print is less than 200MB after selecting the tiny model and the CPU load spikes while the voice prompt is being processed. The built-in Apple Speech model is almost real time. There must be some other difference that is causing FluidVoice to take so long compared to whisper.cpp on the same machine. Any advice? |
Beta Was this translation helpful? Give feedback.
Replies: 3 comments
-
|
1.6.3 has a fix for all whisper models :) hold on! |
Beta Was this translation helpful? Give feedback.
-
|
That seems to have fixed it. Curious, what was the issue? |
Beta Was this translation helpful? Give feedback.
-
|
I wss just lethargic and added a bad backend which was unoptimized. Because
no one used whisper much. Now that’s changing and I also changing the
backend:)
…On Tue, Jul 14, 2026 at 11:38 AM Brenden Soares ***@***.***> wrote:
That seems to have fixed it. Curious, what was the issue?
—
Reply to this email directly, view it on GitHub
<#575?email_source=notifications&email_token=BVSOW2RW7XWA26KXGAEMCMD5EZ437A5CNFSNUABIM5UWIORPF5TWS5BNNB2WEL2ENFZWG5LTONUW63SDN5WW2ZLOOQXTCNZWGM4TINBVUZZGKYLTN5XKY43UMF2GKX3DNBQW4Z3FUVSXMZLOOSWGM33PORSXEX3DNRUWG2Y#discussioncomment-17639445>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/BVSOW2SLMEBU7OKLYPY7R535EZ437AVCNFSNUABJKJSXA33TNF2G64TZHMYTANRRGMZDOMZRGE5UI2LTMN2XG43JN5XDWMJQGQYDONBSHGQXMAQ>
.
You are receiving this because you modified the open/close state.Message
ID: ***@***.***
com>
|
Beta Was this translation helpful? Give feedback.
1.6.3 has a fix for all whisper models :) hold on!