←
Freestyle79 posts
Freestyle avatar

Freestyle

@FreestyleVoice

Track anything and get notified

◷ Joined May 2026▧ 24 Photos and videos
Freestyle avatar
Freestyle@FreestyleVoice·Jul 27

Recent improvements to Freestyle Voice have made dictating medical terms a lot more accurate. New models & post processing has improved Freestyle's ability to use context clues to correct medical terms. It can also format what you say into SOAP notes and emails.

0:54
Matt Wang avatar
⟳ Freestyle retweeted
Matt Wang@mattt8p·Jul 1

I'm super excited to show y’all Freestyle Transcribe, the open source voice dictation pipeline that rivals Wispr Flow’s performance. Check it out

0:44
Freestyle avatar
Freestyle@FreestyleVoice·Jun 29

Sharing how we built an effective voice dictation pipeline. Works incredibly well on capturing terminology devs use. Runs at ~600ms latency, on par with Wispr Flow.

Matt Wang @mattt8p Jun 29 Article Voice dictation pipeline tuned for devs When we first put out Freestyle as an open source project, we wanted to allow developers to configure whatever transcription pipeline they wanted. Choose your own speech model, whether or not you want

Freestyle avatar
Freestyle@FreestyleVoice·Jun 24

Open source Wispr Flow means you can build cool extensions on top of it. And get full customizability.

I wanted to hack into Wispr Flow and customize it however I'd like. Do cool things like setting post-processing guidance and mix & match model providers. I have full control of the transcription pipeline. Check it out:

0:59
Freestyle avatar
Freestyle@FreestyleVoice·Jun 16

Been voice pilled lately. Sharing a couple of ways to use Claude Code + voice dictation to prompt faster!

Matt Wang @mattt8p Jun 16 Article Effective ways to use Claude Code with voice dictation It was the summer of 2023, I'd just graduated and got my computer science degree with an offer. Cursor had just come out a couple of months prior, but for the most part, we were all still hand writing

Matt Wang avatar
⟳ Freestyle retweeted
Matt Wang@mattt8p·Jun 11

I figured out how to use ASR biasing for speech transcription! One of the coolest features that Wispr Flow has is ASR biasing. ASR biasing is a transcription technique that guides the model with hints on how words are spelled, or what phrases are common.

0:35
Matt Wang avatar
⟳ Freestyle retweeted
Matt Wang@mattt8p·Jun 8

Two weeks into building Freestyle, @MathurAditya7 and I learned a ton about building in voice. For context, we've been working on an oss alternative to Wispr Flow. Thought it was going to be a cakewalk, but there are some technical things that are really challenging. I wrote an article on how we built the floating pill thing. 1️⃣ The pill isn't part of the app. It's its own transparent, non-focusable Electron window that floats above everything. That way it doesn't steal your cursor when dictating. 2️⃣ To get the waveform effect, we use getUserMedia API and attach a Web Audio analyser. Split the results of analyzer into 14 bars rendered by frequency intensity. Check out the article! freestylevoice.com/blog/floa…

Post attachment
Freestyle avatar
Freestyle@FreestyleVoice·Jun 2

The current status of Freestyle... "Is PostHog down" ➡️ "Is post-hug down." 1.12s inference Yeah we have a long way to go 😂

Post attachment
Freestyle avatar
Freestyle@FreestyleVoice·Jun 1

Lately, speech-to-text voice dictation apps like @WisprFlow have taken the world by storm. Credit where it's due, it's a genuinely polished product. Low latency, the post-processing is great, the product feels premium. It’s changed how we do work. But a couple things didn't sit right with me. You're paying $12 a month for it. Every sound you make, every word you say and every transcription passes through their servers. It’s a standing risk to your privacy. There's no reason your private thoughts ever have to leave your device. Plenty of open-source dictation tools already exist, but none of them feel as polished as the paid apps, and that gap is what we want to close. Voice dictation is a commodity and should be free and open source. It's an early preview with lots to build, so we'd genuinely love contributors. Freestyle is a technically challenging and ridiculously fun project to work on. We’re looking to grow our community. If this sounds interesting, come build with us! github.com/freestyle-voice/f…

Freestyle avatar
Freestyle@FreestyleVoice·Jun 1

Excited to announce that we're launching Freestyle! Freestyle is the dictation app that converts your voice into text. Hold down a hotkey, talk, and clean text will appear wherever your cursor is. Speak 4X faster than you type. Freestyle is free, open source, and local first, ensuring your dictations are private and never leaves your device. A couple of highlight features: 1️⃣ Transcription post-processing: grammer and punctuation cleanup via post-processing. No more "um, oh, but". 2️⃣ Choose your own models, cloud or local on-device 3️⃣ Dictionary: custom phrase replacements applied after transcription. (ex. "type script" → TypeScript). Try it out today! freestylevoice.com/

0:33