A complete Android keyboard, built around voice.
Type normally or tap the mic where you already write. Choose a local model, or turn voice off and keep the rest of the keyboard.

memo.wav, 5.8 secondsanswered in 0.34 sRemind me to send a contract to Dana on Thursday, and book the small room for 10.
A real response with word timings, from a synthetic voice.

Send the updated contract to Dana by Thursday and book the small room

- voice models in the Android app
- 13
- languages on Loquela Remote
- 25
- transcribe speed on Remote
- 400×*
- recordings Remote keeps
- 0
* Our own tests: server time for 60-minute MP3 and AAC files, not counting upload.
Loquela for Android
Love Loquela? Take it with you. A complete keyboard built around voice, with your choice of speech models and settings for nearly everything.
13 voice models
They run on the phone itself, matched to what it can handle.
A full keyboard
Autocorrect, suggestions, themes, layouts and clipboard tools, with voice or without it.
Free to start
Everyday typing and local voice cost nothing. Full is a one-time unlock.


from openai import OpenAI
client = OpenAI(
base_url="https://api.getloquela.com/v1",
api_key="lqr_...",
)
with open("memo.wav", "rb") as f:
text = client.audio.transcriptions.create(
model="loquela-remote-beta", file=f
).textLoquela Remote
A speech-to-text API and MCP server running on hardware we own, not in a big AI company's cloud. Send a finished recording and get the text back with word timings. The audio stays in memory while it's transcribed, then it's erased.
400× transcribe speed
An hour of audio in under 9 seconds of server time, in our own tests.
English plus 24 languages
For customers in the United States.
Fits what you already use
Works with the OpenAI SDK, and as an MCP server for Claude, ChatGPT and Cursor.
Plans from $9.99 a month for 5,000 minutes, or pay as you go at $0.0035 a minute.
Loquela Desktop
Hold a key, talk, and your words land in whatever app you're in. Then make it yours: pick the model, patch it for your field and tune every setting.
Works in any app
Hold to talk, or tap once for hands-free. The text goes wherever your cursor is.
Every setting exposed
Choose the speech model, tune the engine, teach it your words and give each app its own profile.
Model patching
Add patch packs for your field and set how strongly each one applies.


Power to the people who use it.
Your model, your settings.
Pick the speech model that suits you, then change whatever you like. Loquela doesn't hide the controls.
Local power for you, our own hardware for business.
On your computer and phone, the models run on your device. For businesses, Remote scales on hardware we own, keeps none of the audio it hears, and costs less the more you use, down to $0.0008 a minute.
Already like your note-taking app? Use Loquela inside it: Desktop can serve an OpenAI-compatible API on your own computer, and Remote serves one online.



