Drop your meeting recording into fwip’s Speech to Text tool. On-device AI transcribes it with speaker labels and timestamps. The recording never leaves your machine. No Otter account. No cloud processing. No third party hearing your board discussion, HR meeting, legal consultation, or therapy session.
How to do it
- Open fwip’s Speech to Text tool.
- Drop your meeting recording in — MP3, WAV, M4A, MP4, WebM.
- Select language.
- Hit Transcribe.
- Review the transcript. Edit, copy, or download.
The AI ran on your device. The recording stayed on your device. The transcript was generated on your device.
Who needs this
Lawyers recording client consultations. Attorney-client privilege means the recording shouldn’t be on a third-party server. Ever.
HR departments recording disciplinary meetings, grievances, or terminations. These recordings contain sensitive personal information protected under privacy legislation.
Therapists and psychologists recording sessions for notes. Clinical recordings are subject to strict confidentiality requirements.
Board members recording strategy discussions. Competitive intelligence, M&A conversations, financial forecasts — none of this belongs on someone else’s server.
Journalists recording sources. Source protection is foundational to press freedom. Uploading a source interview to Otter defeats the purpose.
Anyone who recorded something sensitive and needs a transcript without sending the audio elsewhere.
The problem with cloud transcription
Otter, Descript, Rev, Fireflies, HappyScribe — they all upload your audio to their servers. Their privacy policies say they delete it after processing. Some retain it for model training unless you opt out. Some don’t specify.
The question isn’t whether these companies are trustworthy. The question is whether your recording should exist on anyone’s server other than your own device. For sensitive meetings, the answer is usually no.
fwip’s transcription model runs locally via WebAssembly. Zero network transmission of your audio data. You can verify by disconnecting from the internet after the page loads — the tool still works.
Frequently asked questions
Is this as accurate as Otter? For clear recordings with 1–3 speakers, accuracy is comparable (90–95%). Noisy environments, heavy accents, or overlapping speakers reduce accuracy. Cloud tools have larger models and sometimes edge ahead on difficult audio, but the privacy tradeoff may not be worth it.
Can it handle a one-hour meeting? Yes. Most modern laptops handle 60–90 minutes. For longer recordings, the desktop app is more reliable. Very long recordings may take a few minutes to process.
Does it identify who said what? Yes. Speaker diarisation labels different voices as Speaker 1, Speaker 2, etc. It won’t know names, but you can label them yourself in the transcript.
What recording formats work? MP3, WAV, M4A, OGG, FLAC, WebM, and MP4 (extracts the audio track). Zoom recordings (.mp4), Teams recordings, iPhone voice memos — all work.
Can I export the transcript? Yes. Copy to clipboard or download as a text file. Paste into your document, case file, or notes.