One studio, four kinds of recording
The same recorder underneath: everyone captured in full quality on their own computer, uploaded while you talk, stored where you say. What changes is what you do with it.
Record a podcast remotely without the remote sound
Every voice on its own track at full quality, even when a guest’s Wi-Fi gives up halfway. Nothing to install for guests, nothing to re-record for you.
Separate tracks, already aligned
Per person a WAV and an MP4 that start at the moment you pressed Record and share one length.
Guests just click a link
A green room checks their browser, camera, microphone and free space, and asks about headphones, before they join.
A producer’s view
See who’s too quiet, distorting, echoing or falling behind on upload, as it happens.
Interviews with a clear answer for your ethics board
Participants join from a link, each one’s consent is kept with the recording, and the audio stays on your institution’s server.
Data stays where you say
Your server, your bucket, your keys. When you host Tapry yourself, there is no Tapry account that can see the files.
Consent in the record
Each participant agrees to the wording you set for the studio. Every recording comes with a consent record: who agreed, when, and to which text.
Transcripts without a vendor
whisper.cpp runs on your server and transcribes each person’s track on its own, so every line is labelled with who said it.
From call to timeline without the clean-up
Levelled, aligned per-person files and a combined video are made on your server as soon as everyone’s upload is in. They are plain WAV and MP4 files, for the editor you already use.
Even volume by default
Every track is brought toward −16 LUFS with one volume change and no compression, so voices match without touching a fader.
One length, one start
All files start at the moment Record was pressed and are cut to the recording’s exact length.
Subtitles included
SRT and VTT alongside TXT and DOCX, with each speaker named.
Record sensitive conversations in-house
Run Tapry on a server you control. The code is open, the storage is yours, and transcripts are made on the same machine.
Readable, auditable
AGPL-3.0 server, MIT recorder core. Logs hold ids and errors, never media.
Works when the network doesn’t
Guests record locally even on a poor connection; uploads resume when it’s back.
One Compose file
Caddy, Postgres, SeaweedFS, LiveKit, the worker and the web app on a single machine.