Previously, we introduced a ByeType free speech-to-text tool, although it also claims to be able to replace TypelessHowever, after actually testing it, there are still some drawbacks, such as slightly slower conversion speed, occasional language detection errors, etc. It still falls short compared to Typeless. Recently I discovered SayIt and felt it gets much closer! The conversion speed is extremely fast, completing in less than 1 second after you finish speaking (the official claim is within 3 seconds), and its built-in AI It also corrects homophonic typos, redundant words, and other issues. Additionally, the dictionary feature lets you add commonly used terms to significantly improve proper noun recognition accuracy.Mac、Windows All versions are available, and they’re all free to use.
SayIt Free Voice-to-Text Tool Introduction and Tutorial
SayIt is a person named Jackle Chen This free tool uses Whisper for speech transcription, then sends the results to an LLM for organization. Like Typeless, it doesn’t just transcribe voice to text—it automatically removes filler words like “um,” “uh,” and “like,” organizing everything into ready-to-use content. No manual editing needed, which is incredibly useful for writing or work communication. It supports global hotkeys, so you don’t need to switch windows—you can use it directly in any app, whether AI tools, LINE, browsers, or email.
What’s particularly noteworthy is that SayIt’s AI organization feature uses the Groq API. However, there’s no need to worry about the API—Groq is very easy to sign up for. You can obtain your key simply by registering, and no credit card is required. Moreover, the daily free quota is more than enough for most users, so it’s essentially free to use. Even if you exceed the free limit, the costs are minimal (around NT$3.5 per hour of transcription). Compared to subscription-based services like Typeless, which costs NT$300 or more per month, SayIt is significantly cheaper. Additionally, it emphasizes data privacy with local storage—all recordings and history are saved on your own computer and won’t be uploaded to the cloud, giving privacy-conscious users peace of mind.
Main Features
- Press and hold to speak, release to auto-paste the text — takes about 3 seconds
- Whisper Speech-to-Text + AI Automatically Organizes Sentences, Removing Verbal Fillers
- Supports global hotkeys, works in any app
- macOS / Windows dual platform support
- Nearly free (daily Groq quotas handle everyday use), with costs staying extremely low even when exceeding limits.
- Local storage of recordings and logs enhances privacy protection
- Customizable AI prompts and tone styles
- Supports custom dictionaries, enhancing recognition accuracy for proper nouns
Follow the link above to the SayIt website and download the version that matches your system. I’m on macOS (M-series CPU):
Before installing, you can go to the Groq website to apply for an API key first. If you don’t have an account yet, just sign up for free. After logging in, you’ll see the API Keys option in the upper right corner:
Click Create API Key to create:
Enter a name and set whether to auto-delete upon expiration. Default is no expiration date:
Next, you’ll get an API Key. Copy it, as you’ll need to enter it when using Sayit:
For Mac version, enable Accessibility permissions, click to open System Settings:
Just open SayIt and you’re done:
After opening SayIt, scroll down the settings menu to find the Groq API Key setting option. Enter the API Key you just obtained and press Save. Then proceed to set up the model. The speech transcription model is the Whisper series:
There are four text organization models, with the differences marked on the far right. Kimi K2 Instruct is the smartest, but if you need a lot of voice input daily and your free quota runs out easily, you can start with the cheapest option: Llama 4 Scout 17B:
I’m configured with Whisper Large V3 Turbo + Llama 4 Scout 17B:
The AI organization prompt has two presets: “Concise” and “Aggressive”. If you want results closer to Typeless quality, switch to Aggressive, or set your own prompt:
It also has a built-in smart dictionary learning feature that automatically learns new vocabulary through AI analysis.
Custom dictionary: When you encounter a word you frequently say but is consistently misrecognized, you can add it here:
Shortcuts default to holding Fn, but you can also customize them to your preferred keys.
Then you can start talking. A status bar will appear at the top during transcription, and from my testing, as long as the sentences aren’t too long, it transcribes in basically 1 second:
Of course, there’s also a dashboard that tracks how long you’ve dictated, word count, total usage, time saved, and your remaining free quota for today, etc.:
Hover over the status bar below to see how much free quota remains for each item:
There is also a history feature. In addition to automatically saving audio files and transcriptions, it also keeps a record of the text before and after editing.
Currently, a known bug in the Windows version is that voice input doesn’t auto-fill. It will appear if you paste using right-click or Ctrl-V.
Source: KOCPC Chinese