$ dikt --transcribe
"Hello, world"

speak.transcribe.ship.

Speech-to-text for developers who hate typing.
Multiple AI models. 30+ languages. Zero friction.

download v1.0.0 view source

// features

Built for productivity

Shortcut Recording

Ctrl+Space to start. Ctrl+Space to stop. That's it.

Multiple AI Models

Parakeet, SenseVoice, Whisper - choose your engine.

30+ Languages

25 European languages + Asian languages support.

Auto Mute

System audio muted during recording. Clean captures.

// quickstart

Four steps. Done.

Install

Download for your platform. One command install.

Configure

Select your model. Pick your shortcut.

Record

Press shortcut. Speak. Press again.

Paste

Text auto-copied. Paste anywhere.

// models

Choose your engine

modelsizelanguagesspeed
Parakeet V3478 MB25 EuropeanFast
SenseVoice160 MBChinese, Japanese, KoreanVery Fast
Whisper Medium492 MB99 languagesAccurate
Whisper Turbo1.6 GB99 languagesBalanced

model	size	languages	speed
Parakeet V3	478 MB	25 European	Fast
SenseVoice	160 MB	Chinese, Japanese, Korean	Very Fast
Whisper Medium	492 MB	99 languages	Accurate
Whisper Turbo	1.6 GB	99 languages	Balanced

# fedora / rhel / centos

Ready to speak?

Add the repository and install with dnf.

Step 1: Add repository

$ sudo dnf config-manager --add-repo https://rohithmahesh3.github.io/dikt-rpm/dikt.repo

Step 2: Install

$ sudo dnf install ibus-dikt

view source

Updates via sudo dnf upgrade