Pinna detects who-spoke-when and turns your call and meeting audio into speaker-labeled, time-aligned transcripts. Submit by web or API and get back a transcript that tells the voices apart.
Detects how many speakers are in the audio and labels who-spoke-when automatically — no manual tagging, no guessing.
Whisper-class accuracy in Hebrew, Arabic, Russian and English — including the languages cloud tools get wrong. Speaker diarization built in.
Hebrew, Arabic, English, and Russian — including right-to-left scripts. Auto-detected, or force a language when you know it.
Upload in the web app or call the API. Same engine, same speaker-labeled result, whichever you use.
The core runs with no third-party cloud-API dependency — the foundation that makes on-prem and airgap deployment possible (available with Enterprise).
Every speaker turn carries a timestamp, so you can jump straight to the moment in the recording a line came from.
Drag a file into the app, watch the job progress, and download the transcript when it's done. The simplest way to get started.
Do not send recordings by email. All messages are currently refused and do not reach transcription, including messages from allowlisted senders.
Submit jobs programmatically and pull results into your own systems. Built API-first, so automation is a first-class path, not an afterthought.
* Unlimited is subject to fair use.
Every plan includes all languages and diarization. You pay for hours and delivery mode, not features.
Free, Starter and Pro state a monthly allowance in hours. At launch, nothing measures your usage against it. You will not get a reminder, and nothing stops you when you pass it: the allowance describes what the plan is sold as, not something the service currently counts.
There is no enforced ceiling at launch, and no usage tracking behind it. Reaching or passing your included hours does not stop you from submitting more, and no notification is sent as you approach them. Metering and enforcement are being built. When they arrive, this page will say what they do and how the monthly period is counted.
The core diarization and transcription run on a self-contained engine. Your audio doesn't get handed to a third-party transcription API to do the work.
Because the engine is offline-capable, Enterprise can run Pinna entirely on infrastructure you own — including airgapped networks — so audio never leaves your perimeter.
The authenticated app manages API keys and job access, so only authorized users can submit through the app or API and read results. Email submission is unavailable, including for allowlisted senders.