ΕΛ Sign in Try 5 minutes free
For oral history & archives

Speech in. Metadata-enhanced record out.

An oral history interview is evidence — and Akoé is an AI tool for the historians and archivists who keep it. Automatic transcription, each voice labelled, every word timestamped, an editor built for careful checking: what enters the archive is what was said.

Upload an interview
First 5 minutes free. No card, no subscription. After the trial, we charge €5 per hour of audio pro-rated, starting from a €1 minimum charge.
EU-owned infrastructurenot just EU regions of overseas cloudsDeleted on your scheduleaudio and transcript togetherNever used to train modelsno third-party AI touches your filesExact price before you pay€5 per hour of audio, to the second

Word-timed, speaker-labelled, citable

Every word carries its own timestamp, so a quotation can point to the exact moment on the tape — and while you check, the audio plays from the word you're on.

Speakers are separated automatically and named by you. The mentions index collects the people, places, organisations and dates in the testimony — plus labels you define — and exports as CSV: raw material for a finding aid.

When a passage is uncertain, it stays flagged until a human has looked at it. The transcript that leaves Akoé is one a person has checked, not one a model has promised.

The formats archives actually ask for

TEI XML

Structured export for digital archives and text collections — corrections and timings included.

Word & plain text

An interview layout for reading copies and deposit files, or clean text for catalogues and summaries.

Word-timed data

CSV and JSON with per-word timing and speaker labels — for indexes, research datasets and custom pipelines.

Subtitles

SRT and VTT timed to the word, for access copies and listening stations.

Custodianship you can explain to a narrator

Interviewees trust you with their memories; you should be able to say plainly where the files go. Every company that touches the audio is EU-owned and EU-based, publicly listed. Recordings are encrypted at rest, per workspace.

Retention is a schedule you choose — 7, 30 or 90 days, or keep until you delete — and audio and transcript are deleted together. Projects are private to their creator by default; sharing is explicit.

For collections: annual hour blocks on invoice, purchase orders and bank transfer, a data-processing agreement, priority processing — write to billing@toolbox21.com.

Frequently Asked Questions

Fair questions before your first upload

Can I cite a specific moment?
Yes — every word is timestamped, and exports keep the timings. A quotation in an article or exhibition text can point to the precise second of the recording.
We have a large collection of recordings. How do we start?
Uploads resume on their own if a connection drops, long recordings are priced the same flat €5 per hour of audio, and institutions can buy annual hour blocks on invoice with a data-processing agreement. Write to billing@toolbox21.com and tell us about the collection.
What happens to the recordings after transcription?
Whatever you decide: they're deleted together with their transcripts on the schedule you set — 7, 30 or 90 days — or kept until you delete them. Immediate deletion is always available, and nothing is ever used to train models.
Does it handle older or difficult recordings?
Automatic transcription struggles when the recording does — distance, noise, overlap, underrepresented dialects. We won't paper over that with adjectives: the editor is built around the gap — uncertain words stay flagged until you've looked, and the audio plays from the word you're checking.
Does it mark laughter, pauses, or emotion?
No — it produces a pure transcript of speech. Non-speech sounds (laughter, crying, a cough), tone and emotion aren't detected or marked, and silence produces no text. Oral-history conventions differ on how much of that belongs in a transcript; whatever yours require, you add it by hand in the editor, with the audio playing and every word timestamped.
See alsotranscribe interviews with AIGDPR-compliant transcriptionhow to put subtitles on a videotranscribe Greek audio

Your first 5 minutes are free

A real recording or just your voice—see how it works. No card details needed.

Upload a recording Start dictating