AI Research lab focused on on-device dictation and more. Try out FluidVoice at https://nitter.cf/t.co/RRPbdflERx 500K downloads. 12K GitHub stars.
San Francisco
Joined August 2025
- Tweets2.6K
- Following445
- Followers3.6K
- Likes3.9K
Pinned Tweet
We took NVIDIA's Parakeet and made it better.
Introducing Fluid Speech Mini: the best Open Source Model purpose-built for Dictation 🚀
With SOTA performance under 300 MB
Run it on your iPhone, Mac, or PC. No cloud, no internet.
Dictation shouldn't need a server.
Your voice should be yours and we will make sure of that.
Open weights and more info below!
The most productive people have already stopped typing.
An average person types 2,000 words a day.
That is ~1 hour gone.. Just typing.
If you compound it over a year..
That is 6 work weeks just spent typing.
Do yourself a favor, and stop typing.
Spend time with your kids.
Watch a movie with your wife.
Go out with your friends.
FluidVoice retweeted
I didn't know @FluidVoiceApp will give me access to both the Windows and the Android version in just a little span of time.
Currently Beta testing their Android version to write this tweet.
I'm feeling relieved since my words are not being uploaded to cloud anymore.
I have to say if you're still paying for dictation in 2026,
I don't think you're allowed to have opinions about Al, especially Local AI.
We've taken on-device dictation and made it as good as cloud dictation. For free.
What are you even paying for? 👀
Kinda bored of making the best dictation app out there…
so now we're coming for the best ASR speech models ;)
Fluid Speech Pro: #1 on the Open ASR benchmark (9 public sets)
still early - but try it in FluidVoice Beta 👇 and let me know what you think!
oh, and while you're there, play with Fluid 1.1 Splash for formatting and let me know once you've uninstalled other apps 😌
Speech Pro - Open Weights soon.
FluidVoice retweeted
best $0 upgrade i’ve made this year
switched from wispr flow ($15/mo) to fluidvoice (free + open source) a few months ago
- super fast
- works offline
- opus can use the same local model to transcribe my videos, understand what’s happening and know exactly where to cut
haven’t missed wispr flow once
FluidVoice retweeted
This is wildly fast to load now too
Big fan of this. I think I was annoyed by having to swipe back and forth to load the model, but it's actually not that big of a deal when the speed increases, which I think is accomplished here by the smaller model
We took NVIDIA's Parakeet and made it better.
Introducing Fluid Speech Mini: the best Open Source Model purpose-built for Dictation 🚀
With SOTA performance under 300 MB
Run it on your iPhone, Mac, or PC. No cloud, no internet.
Dictation shouldn't need a server.
Your voice should be yours and we will make sure of that.
Open weights and more info below!
FluidVoice retweeted
Update on this: I tried four of the dictation apps people recommended, and ended up choosing @FluidVoiceApp. Easy recommendation to anyone.
It's perfectly what I wanted:
- OSS, lets me pick (and switch) the model.
- Runs completely locally, transparent privacy throughout.
- Easy setup (no nonsense hunting settings or creating accounts)
- Isn't trying to do a bunch of things, just great dictation.
- It's fast and really accurate (Parakeet TDT v2 feels faster and more accurate on my M5 than even Wispr Flow).
Thanks @saadiq @TheBenMeadows and @CyberDuck60 for the recommendation.
FluidVoice retweeted
If you haven’t started using your voice, you’re missing out. Fluidvoice.com
Today we're introducing Fluid Speech Mini - the Best Open-Source speech model for on-device Dictation
Fluid Speech Mini surpasses Whisper Large, and Parakeet in Accuracy in noisy environments,
While being ~10x smaller than Whisper Large, and ~5x smaller than Parakeet.
So you can run it for on-device accurate dictation on your MacBook, or even iPhone 11.
More info on Fluid Speech Mini below!
FluidVoice retweeted
switched over to this model from parakeet last week and it’s been great on my M1 Mac
Parakeet was super fast for dictation, but when I added cleaning, my Mac wasn’t fast enough to process the speech in realtime and it created too much friction.
So I’ve been using it without cleaning and using the raw dictation for months.
Swapped over to fluid speech mini and enabled cleaning and it’s nearly instant, and includes the formatting and cutting out ums and ahs like you get with wisprflow
So, in short, if you’re looking for fully on device text to speech dictation, no subscription costs and it works on older devices, try this instead of wisprflow
Today we're introducing Fluid Speech Mini - the Best Open-Source speech model for on-device Dictation
Fluid Speech Mini surpasses Whisper Large, and Parakeet in Accuracy in noisy environments,
While being ~10x smaller than Whisper Large, and ~5x smaller than Parakeet.
So you can run it for on-device accurate dictation on your MacBook, or even iPhone 11.
More info on Fluid Speech Mini below!
Find me a better dictation app for Windows that doesn't send your Voice to cloud and does better formatting than the cloud models. I'll wait...
Comes with State of the Art formatting models tuned for all your GPUs and CPUs
Early preview link below. Enjoy!
Linux coming very soon ;)
FluidVoice retweeted
FluidVoice on Mac is surprisingly good.
The exact same model launched by someone famous would’ve crossed 100K views overnight.
Sometimes I wonder what’s even the point of spending countless hours researching, engineering, and pushing the frontier forward when the name behind the work matters more than the work itself.
It’s genuinely disheartening and demotivating 😅
Feels like if you’re not already somebody, you’re NGMI in this world of paid promotions and personal brands.
But we keep showing up. We keep building.
Another big launch coming soon :)
Maybe one day the work will speak louder than the name behind it.
FluidVoice retweeted
Anyone out there looking for a Top Tier Dictation app and more. Look no further than Fluid Voice, if your looking for the best forget about the rest. Fast, Accurate and so much more. @FluidVoiceApp
The exact same model launched by someone famous would’ve crossed 100K views overnight.
Sometimes I wonder what’s even the point of spending countless hours researching, engineering, and pushing the frontier forward when the name behind the work matters more than the work itself.
It’s genuinely disheartening and demotivating 😅
Feels like if you’re not already somebody, you’re NGMI in this world of paid promotions and personal brands.
But we keep showing up. We keep building.
Another big launch coming soon :)
Maybe one day the work will speak louder than the name behind it.
Today we're introducing Fluid Speech Mini - the Best Open-Source speech model for on-device Dictation
Fluid Speech Mini surpasses Whisper Large, and Parakeet in Accuracy in noisy environments,
While being ~10x smaller than Whisper Large, and ~5x smaller than Parakeet.
So you can run it for on-device accurate dictation on your MacBook, or even iPhone 11.
More info on Fluid Speech Mini below!
Should we rename our app to FerrariVoice? 🤭
As much i've enjoyed using @superwhisper and bought a lifetime plan, I've spent the past two weeks or so using @FluidVoiceApp and have been absolutely delighted by the experience.
Both apps have kinks that need to be resolved but it feels like @FluidVoiceApp has most of the user experience I'm looking for: it's incredibly fast with almost zero latence (one of my biggest gripes with @superwhisper), keeps your copy on your clipboard, transcription intelligence is insane, the custom dictionary is awesome and it has tons of other features I haven't even looked into yet.
Most important of all, it's open-source and private unlike @WisprFlow which I got rid of with the stunt they pulled with people's data a few months ago.
It really feels like the Ferrari of dictation apps
