LESS TYPING. MORE THINKING.
Give your
thoughts a voice.
Just say the words as they come to you.
Check them in a small window, then send them where you usually type.
Melo is a Mac voice-input app that keeps pace with the way you think.
macOS 26.2+ · Japanese / English · Coming soon
MEET YOUR OTHER KEYBOARD.
Stop, then check and edit right there. Send the words you’re happy with.
A faithful recreation of the current app’s 300×220 overlay. The background document is an illustration.
This demo performs no recording, speech recognition, or clipboard actions. Speed and accuracy vary by environment.
01 / KEEP YOUR TRAIN OF THOUGHT
Try saying it
before typing it.
A quick email reply, an idea that just surfaced.
Capture them at the speed of speech.
Start recording with the hotkey, press it again to stop.
Preview partial text while recording, and edit in place after it ends. Everything from dictation to pasting stays in one small window.
02 / SMALL WINDOW. YOUR RHYTHM.
On screen, only
what you need.
Melo lives in your Mac’s menu bar.
The small overlay appears only while you speak.
Choose from four waveform styles — whichever is easiest to read.
Display position and hotkeys can be changed in Settings.
Using the overlayFIND YOUR RHYTHM
Rendered examples following the current app’s drawing logic.
03 / YOUR WORDS, SPELLED YOUR WAY
Frequent terms,
in your own spelling.
Names, specialist terms, everyday abbreviations.
Register a “reading” and an “output” in the user dictionary,
and after transcription the text is rewritten in the form you prefer.
明日の資料に、AIの活用例を追加。
Example · longest-match dictionary replacement
04 / CLOSE TO YOU. ON YOUR MAC.
The AI that turns voice into text
lives on your Mac.
Speech recognition runs on local Whisper.
Transcription and dictionary conversion happen on your Mac.
An AI that polishes the text can be added when you need it.
Use it locallyAfter a one-time setup of the required models, Melo runs in a local configuration. Downloading the models does require network access.
Choose where formatting goesOptional AI formatting works with local OpenAI-compatible endpoints or cloud providers. Enabling cloud formatting sends your transcript to the configured destination.
A LITTLE SETUP. A LOT MORE FLOW.
Voice-powered input,
on your Mac.
From first-time setup to hotkeys and the dictionary.
The guide walks you through, one step at a time.
- 01
Prepare a compatible Mac
The current app requires macOS 26.2 or later. Also set up the Whisper runtime and models needed for speech recognition.
- 02
Grant the required permissions
Recording uses the microphone permission; direct input into other apps uses the accessibility permission.
- 03
Press, speak, check, deliver.
The default hotkey is Option+Space. Choose toggle mode (press to start and stop) or push-to-talk (record only while held).
GOOD TO KNOW
Common questions.
Do I need an internet connection?
After preparing the required runtimes and models such as Whisper, speech recognition works in a local configuration. If you enable cloud formatting, your transcript is sent to the configured destination. Downloading the models does require network access.
Can I mix Japanese and English?
Recognition modes include auto-detect, Japanese, and English. Accuracy and processing time vary with the model, your recording environment, and how you speak.
Does this page use my keyboard or microphone?
No. This page is a visual demo using authored examples. It records no audio, and it changes no app settings or other apps’ input.
What do I need to get started?
The current release candidate requires macOS 26.2 or later. You will also need the microphone permission plus the Whisper runtime and models. Direct input additionally requires the accessibility permission.
Where can I download it?
Coming soon. The download link will be posted here as soon as it is ready. In the meantime, the setup guide lists everything you will need.
GIVE YOUR THOUGHTS A VOICE.
Start your next idea
with your voice.
Voice into text. Text into your next step.
See how Melo worksComing soon · macOS 26.2+ · Japanese / English