Open-Source Dictation Is Here… Goodbye Subscriptions

BBetter Stack
컴퓨터/소프트웨어가전제품/카메라AI/미래기술

스크립트

00:00:00Just to talk to your computer instead of typing it all out, you end up with another subscription
00:00:04to something like Whisperflow. And yeah, it's good. You'd think the free thing we already have
00:00:10on Mac would be good enough, but it's not. This is Fluid Voice. It runs completely on your own
00:00:17machine and it's growing in popularity as a strong competitor to Whisperflow and all the others.
00:00:27Fluid Voice is a free open source Mac app that turns your voice into text. And every part of that
00:00:33happens on your own computer. Nothing ever gets shipped off to a server somewhere, which is a huge
00:00:38win. Two reasons this is good for us. And there are the exact two things devs complain about. First,
00:00:44your voice and whatever you're dictating might be code or private, and this never leaves your Mac.
00:00:50Second, you're not paying another monthly bill to do it. Here's how simple this is.
00:00:55A model called Parakeet listens and writes down what you say. Then a second local model called
00:01:02Fluid Intelligence acts like an editor sitting right next to you, fixing your capitalization,
00:01:07punctuation, and structure as you talk. You hit one hotkey that you set, speak, and the cleaned up text
00:01:14drops into whatever your cursor is already at. Into cursor, clods, slack, mails, notes. It doesn't
00:01:19really care which one. Now, this is super simple to spin up, so I'll keep the demo short too.
00:01:24It's a one-line install. You open it, give it mic and accessibility permissions. Every Mac app
00:01:31makes you do this, so you have to do it anyways. And then you can choose a hotkey or use the one
00:01:36built in. So there's no real difference here. Now, watch this. I hold the key and I just talk normally
00:01:42here. Can you check over my code and make sure it's set up with some good practices to prevent
00:01:47bugs. And it didn't just collapse. Look at what it just did. It capitalized the line, punctuated it,
00:01:55kept my technical word spelled right, and formatted it like an actual comment. I didn't fix one thing.
00:02:01No spinner, no waiting, because there's no cloud to actually wait on here. So where does this actually
00:02:07sit next to everything else? Well, this is actually four times faster than using other ones already out
00:02:12there. And on top of that, this is where it gets interesting. The dictation already on your Mac is
00:02:16free and for a text message, it's fine. For a commit message or a doc, someone actually going to read
00:02:22that, it's probably not so great. Whisperflow is the paid, so you'd expect it just wins here. But
00:02:27it's subscription. Your audio goes to a cloud. For a lot of us, that second part is a deal breaker
00:02:33before we even get to the price. Super Whisper and the rest are good tools. Still cost money.
00:02:38They don't quite match the local cleanup. Fluid voice is actually doing. So the free one that was
00:02:43supposed to be the weakest, it's the only one that's free open source and does the smart formatting
00:02:48right there on our machine. On Apple Silicon, it's fast enough that you can forget it's even thinking.
00:02:55Now, before you run off and install this, it's great that it's free, right? There's no subscription.
00:02:59It's great that it's private. Our voice stays on our Mac. On M series machines, it's fast and the
00:03:05formatting is actually pretty good. And it's open source and actively worked on, so problems are
00:03:10getting fixed. But on the flip side of things, it's Mac only right now. iOS and Windows are on the
00:03:16wait list. So if you live on Windows, this isn't your tool just yet. The editor model is about three
00:03:22and a half gigs. So it's a real download. It's not a quick one. It's best, like I said, on the
00:03:27silicon chips, it does run on Intel. It's just slower. And if you dictate in a language other
00:03:32than English, you might have to tune it to get it just right. But you should know all that now,
00:03:36not after you spend 10 minutes trying to download this the hard way. You probably expect me to say
00:03:42it's worth it now, but I won't. For most Mac devs, yeah, okay, this might be worth giving a shot.
00:03:48It's worth it if you don't want another monthly payment just to dictate things for you. Which,
00:03:52come on, who does want that? Use it for boring typing, email, docs, code comments,
00:03:57Slack messages, heck, talk to Claude. On a Mac M4 Pro, it's really nice because it's all local and it's
00:04:03quick. If you need Windows or iOS right now, sorry, but this can't help you. So here's what you could
00:04:08actually do. Run it as your everyday tool on Mac and keep something cross-platform around for when
00:04:13you're not on Mac. You get the best of both worlds and it costs you nothing if you're using Mac.
00:04:18The GitHub link and the brew command are both in the description below. If you enjoy coding tips and
00:04:23tricks like this, be sure to subscribe to the BetterStack channel. We'll see you in another video.

핵심 요약

Fluid Voice provides a free, private, and high-speed alternative to subscription-based dictation services by processing all audio and formatting locally on Apple Silicon Macs.

하이라이트

  • Fluid Voice is an open-source Mac application that performs speech-to-text dictation entirely on the user's local machine without cloud server usage.

  • The application utilizes a two-model system: Parakeet for transcription and Fluid Intelligence for real-time formatting, punctuation, and capitalization.

  • Processing occurs locally, ensuring that sensitive data such as code or private communications remains on the user's device.

  • Fluid Voice performs dictation 4x faster than other currently available transcription tools.

  • The editor model requires a 3.5 GB download and performs best on Apple Silicon, though it functions on Intel-based Macs with slower results.

  • Fluid Voice is currently compatible exclusively with macOS, with Windows and iOS versions currently pending on a waitlist.

타임라인

Local Privacy and Cost Advantages

  • Fluid Voice performs all speech-to-text processing locally, keeping data off external servers.
  • Users avoid monthly subscription fees associated with cloud-based dictation services like Whisperflow.
  • Local processing protects sensitive information such as source code and private documents.

Standard dictation software often relies on subscription models and cloud processing, which raises security concerns for developers and power users. Fluid Voice eliminates these issues by running entirely on the local machine. This ensures that personal or sensitive dictation never leaves the hardware, offering a secure, cost-free alternative.

Technical Architecture and Implementation

  • A local model named Parakeet handles initial transcription of spoken audio.
  • A secondary model, Fluid Intelligence, automatically handles capitalization, punctuation, and structural formatting.
  • The application integrates via a user-defined hotkey, injecting text into any active cursor location.
  • Installation requires only a one-line command and basic system permission granting.

The system utilizes a dual-model approach to ensure dictation accuracy and output quality. Once the user triggers the pre-set hotkey, Parakeet captures the audio while Fluid Intelligence simultaneously polishes the text. This process happens in real-time without cloud lag, allowing formatted text to appear instantly in applications like Slack, email clients, or code editors.

Comparative Performance and Limitations

  • Fluid Voice delivers 4x faster performance than competing transcription tools.
  • The tool provides superior local smart formatting compared to existing free Mac dictation options.
  • Hardware requirements include a 3.5 GB model download and preference for Apple Silicon chips for optimal speed.
  • Current platform support is limited to macOS, with other operating systems remaining on a waitlist.

Performance benchmarks indicate that Fluid Voice exceeds the speed of established transcription alternatives. While the software excels on Apple Silicon, Intel Macs experience slower execution. Users interested in the tool should be aware of the significant initial download size and the current lack of Windows or iOS compatibility, necessitating a hybrid approach for cross-platform users.

커뮤니티 글

모든 글 보기