Dictate Keyboard (Whisper AI)

Dictate turns your voice into text — right from your keyboard. Speak instead of type in any app and let a powerful speech-recognition AI do the rest. It supports all common languages and even adds punctuation for you, so you never have to dictate it. Talk naturally, talk fast, or even sing — and it gets transcribed correctly.

Dictate is a full keyboard in every sense: dictate, or type with glide typing, word suggestions and an autocorrect that reads where your fingers land — no switching to another app or keyboard required.

Highlights:
- 🎙️ Reliable speech-to-text powered by Whisper and other top transcription models
- 👆 Push to talk: hold the mic and speak, let go and it's sent — slide left to discard, drag up to lock
- ⚡ Real-time transcription: your words appear live as you speak (OpenAI, Deepgram, Soniox, AssemblyAI, ElevenLabs)
- 🔊 System-wide voice input: use Dictate from other keyboards and apps — their mic key transcribes through Dictate
- 📴 On-device transcription: dictate fully offline with Whisper, NVIDIA Parakeet, Canary or SenseVoice — nothing leaves your phone, and streaming models write while you speak
- ☁️ Or buy prepaid credit instead of an API key: nothing you say or write is stored on the way through
- ✂️ Cheaper uploads: long pauses are trimmed and recordings can be sped up before sending — providers bill by length
- 🗂️ Transcription history: re-insert, replay, re-transcribe or pin any past dictation
- ⏱️ Long-form dictation: speak as long as you like — transcribed in the background, cut at finished thoughts, no length limit
- ⌨️ Glide typing, suggestions, spell check and autocorrect in over 40 languages, plus next-word prediction
- 🀄 Chinese input with Pinyin and a candidate row, alongside the Zhengma method
- ⏺️ Classic keyboard-free layout from Dictate 3 — locked in or a swipe away
- ⌚ Wear OS keyboard: dictate from your watch — via your phone or fully on its own
- 📱 Floating dictation button: dictate into any app even when another keyboard is active — six designs, and hold it for a freeform voice command
- 🔄 Many providers: OpenAI, Gemini, Groq, OpenRouter, Mistral, Anthropic, Soniox, Deepgram, AssemblyAI, ElevenLabs, Ollama (local) and more, or your own server — each with its own API key
- 💡 Custom rewording with GPT-5 and many models: translate, summarize or restyle, plus a community prompt library to browse and share
- 😀 Emoji search in 51 languages, GIF search, and settings search to find any option by name
- 📊 Statistics, custom vocabulary and file transcription
- 🌗 Dark and light mode, a high-contrast E-Reader theme and custom accent colors

Privacy first: there is no tracking, and your audio goes only to the provider you choose. With your own API key, your own server or an on-device model, Dictate never talks to us at all. If you pick the optional prepaid credit instead, nothing you say or write is stored on the way through — and that server's source is on GitHub too, so you can check rather than trust.

Accessibility Service notice: Dictate uses Android's Accessibility Service only for the optional floating dictation button. When you turn it on, the service detects when a text field is focused in any app so the button can appear there, and inserts your dictated text at the cursor. It does not collect, store or share any other screen content. This feature is optional and turned off by default.

Getting started, pick what suits you: your own API key (OpenAI, Groq — usually a few cents a month), a model on your device that needs no account, your own server, or prepaid credit. The setup explains each one step by step.

Dictate is completely open source. You can find the source code on GitHub:
https://github.com/DevEmperor/DictateKeyboard

Found a bug or have an idea for improvement? Just send me an e-mail or open an issue on GitHub.

If you enjoy Dictate, I would really appreciate a donation via PayPal — you can find my PayPal link on my GitHub page. 😇

Have fun with Dictate! 😊