Bark
LIVEVoice-to-text that runs on your GPU
Replaces: Whisper Flow, Otter.ai, Whisper API costs
Bark turns your GPU into a local dictation engine. Hold down a trigger key, speak naturally, and the transcribed text is pasted wherever your cursor is. It uses OpenAI's Whisper model running locally on your hardware, so nothing leaves your machine. With CUDA on an RTX 4090, latency is around 0.2 seconds. On Apple Silicon with Metal, about 0.5 seconds. It auto-detects 90+ languages, strips filler words, catches Whisper hallucinations, and works in any app via clipboard paste.
DesktopAIGPUFreeWindows & macOS
CAPS LOCK
READYWhat it does
- >Hold-to-dictate with configurable trigger key (Caps Lock on Windows, Right Option on Mac)
- >GPU-accelerated: ~0.2s on CUDA, ~0.5s on Apple Silicon
- >Pre-buffer captures audio before you press the key, so first words are never cut off
- >Auto-stop detects silence and ends recording automatically
- >Works in any app via clipboard paste (restores your clipboard after)
- >Auto-detects Swedish, English, and 90+ other languages
- >Strips filler words and detects Whisper hallucinations
- >Floating overlay with live waveform during recording
- >System tray with full settings menu
- >Audio feedback (subtle blips on start/done)
System requirements
- -Windows 10/11 or macOS 12+
- -NVIDIA GPU with CUDA 12.x (Windows) or Apple Silicon (Mac)
- -4GB+ VRAM recommended
- -~2GB disk space for Whisper model
$ bark --status
Bark v1.4.0
Model: medium (cached)
GPU: CUDA (RTX 4090)
Trigger: Caps Lock (hold)
[Recording] ▎▍▌▋▊▉█▉▊▋▌▍▎
"Skicka rapporten till Erik"
Latency: 0.19s
How it compares
Feature
BarkFREE
Whisper Flow$8/mo
Price
Free forever
$8/mo
Runs locally
Yes, on your GPU
Yes, local
Latency
~0.2s (CUDA)
~1-2s
Privacy
Nothing leaves your machine
Local (but license server)
Languages
90+ (auto-detect)
90+ (Whisper)
Works offline
Yes, fully
No (license check)
Works in any app
Yes (clipboard paste)
Yes (clipboard paste)
Pre-buffer audio
Yes, never miss first words
No
Hallucination filter
Built-in
No
Open source
Yes
No, closed source
Requires account
No
Yes
Price
BarkFree forever
Whisper Flow$8/mo
Runs locally
BarkYes, on your GPU
Whisper FlowYes, local
Latency
Bark~0.2s (CUDA)
Whisper Flow~1-2s
Privacy
BarkNothing leaves your machine
Whisper FlowLocal (but license server)
Languages
Bark90+ (auto-detect)
Whisper Flow90+ (Whisper)
Works offline
BarkYes, fully
Whisper FlowNo (license check)
Works in any app
BarkYes (clipboard paste)
Whisper FlowYes (clipboard paste)
Pre-buffer audio
BarkYes, never miss first words
Whisper FlowNo
Hallucination filter
BarkBuilt-in
Whisper FlowNo
Open source
BarkYes
Whisper FlowNo, closed source
Requires account
BarkNo
Whisper FlowYes
FAQ
Frequently Asked Questions
Need something custom?
These tools are a side effect of solving real problems for real clients. If your workflow is broken, we can probably fix it.
TALK TO US