📝 Realtime & Offline Transcription

inFin supports on-device transcription without an internet connection, using supported language models. Online high-accuracy transcription requires a connection and uses the allowance of your membership plan. The two modes have different language coverage and capabilities.

Transcription turns what you say into editable, searchable, copyable text. This is what sets inFin apart from an ordinary voice recorder.

What It Does

• Text as you speak: words appear at the same pace as your voice;
• Chinese and English handled together: mixed Chinese-English speech is correctly recognized and punctuated;
• Offline-capable: the on-device model runs on your device; the network being down doesn't stop it;
• On-device Chinese-English transcription on supported devices: this is where we put the most effort;
• Auto cleanup: filler words and homophone typos are corrected on the fly, and paragraphs are split for you;
• Original transcript on hand: the cleaned text goes into the note body; the original transcript can be expanded at any time to check.

Why "On-Device" Matters

Sending audio to the cloud has three problems:

• A bad network will stall it, or even fail;
• Audio leaves your device, which is a privacy concern;
• Long audio takes a long time to wait for, and creates high per-hour costs, making truly unlimited recording impossible.

inFin runs the model on supported devices. Prepare the required languages and models in advance to transcribe offline. Speed and results depend on the device, language and recording conditions.

How to Use It

1. Once recording starts, transcription begins automatically — no extra toggle required;
2. While recording, you can switch the transcription language / translation target language;
3. When recording ends, the text is automatically added to the note body;
4. If you're not happy, tap "Original" to see the unprocessed version, or re-run a high-accuracy transcription (requires membership).

Four Ways to Improve Accuracy

1. Get closer: keep the device within 1 meter of the speaker;
2. Don't cover it: when hand-held, make sure you're not blocking the microphone;
3. Pick the right language: explicitly choosing the main language is more accurate than "Auto";
4. Speak clearly: speak clearly and loudly for better precision.

Our Principle: Only Fix Typos, Don't Rewrite

inFin's transcription only does two things: fix typos, remove filler words.

No over-polishing, no rewriting on your behalf.

Because that's the trace of how you spoke in that moment — we don't touch it.

FAQ

Q: Which languages are supported? Chinese (Simplified / Traditional) and English are the primary ones. Recognition and translation for Cantonese, Arabic, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Malay, Portuguese, Spanish, Thai, Turkish, Vietnamese, Filipino, and more languages are continuously being added. See Realtime Translation.

Q: How much does offline transcription differ from online high-accuracy transcription? Offline is faster, more private, can be used without limit, and works without a network; online offers higher accuracy in professional and complex scenarios, with support for identifying and separating different speakers.