VOICE INPUT
Dictate into any app, with text that actually comes out right.
One shortcut, any app. Code comments, issue descriptions, PR summaries — just say it. Accurate enough that you rarely need to fix anything.
Japanese-specialized model — built to get kanji and proper nouns right
Japanese voice input,
finally accurate enough to use.
You tried voice input. The accuracy was off, so you went back to typing. Sonophie is built around one goal: getting kanji and proper nouns right, every time. Speak, get clean text, paste it anywhere on your Mac. The same accuracy carries over to file transcription and live meeting notes too.
Already have an account? Sign inTHREE MODES, ONE WORKFLOW
You gave up on voice input because of accuracy. That changes now.
Wrong kanji, misread company names, technical terms turned to gibberish. That frustration is why Sonophie exists. Speak, get usable text, paste it — with the least possible editing.
FILE TRANSCRIPTION
The same accuracy, applied to recordings.
A 30-minute audio file processes in under 10 seconds. Meeting recordings, interviews, onboarding videos — fewer errors means less cleanup work after.
LIVE MEETING
The same accuracy, in real-time during meetings.
Works with Zoom, Meet, and Teams without configuration. Accurate transcription means no more "what did they say?" — and no post-meeting cleanup.
PRIVACY FIRST
Sales calls, interviews, internal meetings — all safe to record.
LOCAL STORAGE
Processed in the cloud, never stored there
Audio is sent for AI processing but not retained server-side. Transcripts and audio files live on your Mac only. Nothing accumulates in cloud storage.
NO AI TRAINING
Your conversations never train AI models
Your audio and transcripts are never used to train or improve AI models. Sensitive conversations — hiring, sales, internal reviews — are safe to run through Sonophie.
FEATURES
How Sonophie pushes accuracy even further.
CUSTOM DICTIONARY
Company names and product names, right every time.
Register company names, product names, technical terms, and people's names on top of the Japanese-specialized model. Once saved, they apply automatically to every future transcript — proper noun errors drop to near zero.
FILLER REMOVAL
Hesitations disappear. Clean text comes out.
Even with accurate transcription, spoken filler words remain. Sonophie strips them automatically so the result is clean enough to paste straight into Slack or a doc without editing.
PROMPT-BASED FORMATTING
"Make it a summary" or "3 bullet points" — done.
Specify the output format with a prompt. Meeting notes, to-do lists, email drafts — the manual formatting step you do every time gets automated on top of the clean transcript.
OFFLINE MODE
Keep company data off the cloud entirely.
Offline mode runs entirely on-device — no cloud processing, no data leaving the machine. Use it when corporate policy restricts cloud services, or when there is no internet connection at all.
FULLY OFFLINE
Works without Wi-Fi
On flights, traveling, in network-restricted offices. AI transcription completes on your Mac without an internet connection.
ON-DEVICE MODEL
Data never leaves the device
With on-device processing, audio and text stay on your machine. Zero data sent to external cloud services.
FLEXIBLE DEPLOYMENT
Switch between cloud and local as needed
Cloud models for everyday accuracy, local processing for sensitive sessions or no-internet situations.
FAQ
Frequently asked questions
How is the Japanese accuracy different from other tools?
Sonophie uses a Japanese-specialized model focused on kanji and proper noun accuracy. Company names, personal names, and technical terms that generic Whisper-based tools typically misread are handled more reliably. Combine that with custom dictionary entries and misrecognitions drop significantly.
Can I reformat transcripts with an LLM after transcription?
Yes. You can specify a prompt to transform the transcript into any format — meeting notes, summaries, to-do lists, email drafts. The formatting step you do manually every time gets automated.
How fast can Sonophie process a 30-minute video file?
As a benchmark, a 30-minute video typically processes in under 10 seconds. Speed varies with network conditions and audio quality, but long files are designed not to keep you waiting.
Can it really work offline?
Yes. Sonophie runs whisper.cpp directly on your Mac, so no internet connection is needed in offline mode. Audio stays on the device.
Does it support any file format and length?
Yes. It handles audio and video files without constraints on format or length. The same flow works for short voice memos and multi-hour recordings.
Are there limits on meeting tools or destination apps?
None. Sonophie does not depend on a specific meeting tool or app. It works across Google Meet, Microsoft Teams, Zoom, and general app voice input.
Is my data stored anywhere?
Audio and transcripts are sent for processing but not retained by design. Output stays on your machine. Neosophie does not store customer data.
Do you offer an API or model licensing?
We can accommodate this in some cases. If you need to run everything on your own servers or integrate the model into an existing system, please reach out via the contact page.
Does it work on iPhone, Android, or Windows?
Currently macOS only. iOS, Android, and Windows versions are in development.
PRICING
Every feature, free during beta.
No usage limits, no paywalled features. Your custom dictionary grows the more you use it — worth starting early.
BETA
¥0
Beta — free, no limits
- Unlimited file transcription
- Real-time meetings and notes
- Voice input across apps
- Automatic filler removal
- Prompt-based formatting and custom dictionary
- Offline mode