Welcome to Blabla Radar
Blabla Radar is a free online tool that helps you detect and count the words you say in real-time. Whether you want to analyze your speech habits, track specific words, or count profanity & swear words, our tool is here to assist you!
Initially designed to monitor the use of curse words at home, Blabla Radar has evolved into a versatile application. It can help reduce filler words, improve your public speaking skills, or simply serve as a fun game to play with friends and children.
Main Features
- Add your own words or phrases to track and analyze.
- Real-time word detection using the Web Speech API.
- Compatible with Chrome, Safari, and Edge. Firefox works only on detectors that offer the local model (beta).
- On select detectors (e.g., Gen Z Slang, Swear Jar), an optional local model (beta) runs speech recognition entirely in your browser — no beep, no *** censoring.
- No registration is required, and your privacy is respected—provided you trust your browser's transcription service (e.g., Chrome, Edge, or Safari), or use the local model, which keeps audio on your device.
Technical Limitations
Blabla Radar relies on the Web Speech API, which comes with some browser-specific limitations:
- Chrome on Android: Profanity is often censored (e.g., "shitty" becomes "s*****"). Censored words are still counted, and the local model (beta) does not censor at all.
- Microsoft Edge: Profanity appears as asterisks ("***"), but you can track them by adding "***" as a word.
- Firefox: Firefox does not support the Web Speech API; only detectors offering the local model (beta) work there.
- Mobile Devices: The microphone cannot run in the background, which is not a limitation on desktop devices.
- Activation/Deactivation Sounds: On some devices (e.g., Android), a beep sound is played when the microphone activates or deactivates. This can be silenced by lowering the notification volume — or avoided entirely with the local model (beta), which keeps the microphone open continuously.
Local Model (Beta) & Credits
Some detectors offer a “Local model (beta)” option that runs speech recognition entirely in your browser: a ~60 MB model is downloaded once, cached, and processes your voice on your device — your audio never leaves it. It avoids the Android microphone beep and does not censor profanity.
This feature is built on open-source software: the Moonshine speech-to-text model and MoonshineJS library by Moonshine AI (English model and code under the MIT license), with ONNX Runtime Web and the Silero voice activity detector. Charts are rendered with ApexCharts (MIT).
Privacy Commitment
Blabla Radar does not store any audio or user data. Speech recognition is processed via the Web Speech API, which relies on the transcription services of your browser provider (e.g., Google for Chrome, Apple for Safari, or Microsoft for Edge). While these services process audio to generate transcriptions, no data is sent to our servers. All processing happens between your device and the browser's speech recognition service.
When the local model (beta) is enabled, recognition runs entirely on your device: your audio is not sent to any server at all — not your browser vendor's, and not ours.
The transcriptions are stored for your convenience in your browser and are not accessible to us or anyone else. You can clear all transcriptions at any time using the reset button, ensuring that no data remains in your browser after use.
How Users Find Us
Blabla Radar is designed to help with the following:
- Counting words or phrases while speaking — try the main tool with your own list.
- Tracking filler words like "literally," "like," "actually", and more.
- Detecting curse words in real time — the Swear Jar even adds 50 cents per curse.
- Counting Gen Z slang and brainrot words as they're spoken.
- A free app to analyze spoken language patterns — browse all detectors.