muesli

own your speech; open source; dictations, meeting notes and computer use all powered by on device models for macOS and iOS

maintained by @pHequals7

[muesli @fastspeech2text]

🚨muesli v0.8.2 is OUT!

major improvements across
> hotkey mic activation startup time
> mic failures mid meetings
> new unified timeline across dictations and meetings
> streamlined model download manager
> Apple's native STT model for macOS 26!

go DOWNLOAD/UPDATE🚨

muesli dot works.

[muesli @fastspeech2text]

muesli v0.8.1 is out now for everyone to update on macOS!

plenty of reliability fixes have been shipped along with support for the latest class of GPT 5.6 models!

[muesli @fastspeech2text]

@kn_neeraj1 @kingofknowwhere thank you neeraj - hopefully will merge it by today!

[muesli @fastspeech2text]

@MrAurelian @WisprFlow try muesli if you're on a mac?

we are local, fast and free!

https://t.co/Z4nnez6bqt

[muesli @fastspeech2text]

Very very excited for folks to try this feature out in the next update!

[muesli @fastspeech2text]

@kn_neeraj1 @kingofknowwhere haha that is fair but we do try to close important PRs

[muesli @fastspeech2text]

@kingofknowwhere 3. is possible via OR for a global logic but the latency is not the greatest due to the roundtrip

app specific transforms can be routed via this - will ship an update on this

[muesli @fastspeech2text]

@kn_neeraj1 @kingofknowwhere possible to raise a PR for this instead - will merge it on priority!

[muesli @fastspeech2text]

@krishnangopal_ extremely alpha build

https://t.co/JvbnE2B1RX

[muesli @fastspeech2text]

⭐️⭐️⭐️⭐️⭐️

[muesli @fastspeech2text]

music credits:

https://t.co/IRtsrATNkd

[muesli @fastspeech2text]

break free from subscriptions

https://t.co/5AArAcEzjk

[muesli @fastspeech2text]

testing background CUA the codex way.

[muesli @fastspeech2text]

background CUA is awesome when done well like this?

[muesli @fastspeech2text]

@boisechampion @myles_snider would love to know what did not work for you with the alternatives?

disclosure: we are one but we also do your meeting notes

https://t.co/Z4nnez6bqt

[muesli @fastspeech2text]

welp! we hit our july usage numbers 10 days in advance this month

looking for privacy first telemetry providers that have generous OSS plans

[muesli @fastspeech2text]

july has been a great month for muesli!

thanks for all the love and support!

[muesli @fastspeech2text]

@pizzapurist 🫡🫡

[muesli @fastspeech2text]

it's amazing to see muesli users go the distance to run the entire meeting transcription note generation pipeline on device!

[muesli @fastspeech2text]

@sidin nothing like using your voice to quite literally creating your writing voice!

[www.sidin.co @sidin]

Super quick little note on one writing topic that came up during my jaunt across Bangalore last week: How do I develop a writing voice if I have never really written before?

1. Usually this is not true. You could pass a bunch of emails through an LLM and ask it to parse your writing style.
2. But ok. Let's assume you haven't written much. What next?
3. Get an audio transcription thing like @AiAudiopen , @fastspeech2text or some such. Something that works on your phone.
4. Now go on a walk or something and answer a question that is relevant. Make up something: How does our backend work? Why did I start this company? What does a good hire look like to me?
5. Whatever you like. But just hit record and speak and speak. Forget about the sequencing, or plot, or whatever. Just speak like you are recording a note to self.
6. Now ask the tool to transcribe it with minimal rewrites. Just remove disfluencies, and correct spelling and so on.
7. Save this.
8. Repeat after a few days.
9. Once you have a few such transcripts, compare them. Chances are you will find patterns that hint at your personal writing style. Because often this is just an extension of your thinking style, which is easily mirrored in your speaking/musing style.
10. Now if monologuing solo is a problem, then convince someone at work to sit across you and ask you the question on a Google Meet that is being recorded.
11. Again give your transcript the mildest possible rewrite. So it retains most of your original verbiage.

(If you already have many meetings in Granola or whatever, why not ask your AI thing to look at a selection of your speaking parts, and figure out a writing style?)

Do this a few times, read your output, and chances are you will find a good "voice". And now you have something to work with.

The cool thing IMO about many speech-to-text things is that they rarely hallucinate, and will mostly stick to what you said. All you need to do is to invest a little bit of time yapping.

[muesli @fastspeech2text]

@JohnSlagboom1 not surprising considering this is what happens when you offload STT to the cloud

try muesli? we're local, fast and open sourced!

https://t.co/Z4nnez6bqt

[muesli @fastspeech2text]

@Jack_Timonen @WisprFlow would love for you try muesli

it's entirely local, fast and OSS

P.S we can do your meeting notes as well

https://t.co/Z4nnez6bqt