Guides, articles, experience reports.

Record your meeting (any audio or video format, up to 2.5 GB), then upload it to the platform. The transcript, speaker identification and first structured draft of the minutes are ready in under an hour. Review, correct if needed, validate the draft → review → published workflow, then export to Word with your official template — ready to sign.

No. Infrastructure is hosted in a Strasbourg data centre under French jurisdiction, with strict per-organisation isolation (a dedicated PostgreSQL schema). No transfers outside the EU take place by default. For ANSSI / SecNumCloud requirements, deployment on Outscale or OVHcloud SecNumCloud is available on a project basis. For organisations with even stricter security policies, on-premise installation within your own infrastructure remains possible.

SaaS is hosted by us in France: immediate start, no installation, automatic updates, support and SLA included. On-premise is installed within your own infrastructure through Docker: no cloud dependencies, no outbound data, and you manage updates yourself. Both modes share exactly the same product.

The architecture is designed to facilitate your compliance — GDPR obligations also depend on your own processing activities. A signable Data Processing Agreement (DPA) is supplied, the processing register is integrated, and access / rectification / erasure / portability rights are available in the user interface (Settings → Portability). The full compliance pack is publicly published at client.transcribia.eu/compliance.

Up to 97% on a good-quality recording (microphone close by, little background noise, speakers not talking over one another). Accuracy drops with noise, overlap and distance from the microphone. To optimise results, use a quality central microphone and add your specific vocabulary to the sector glossary (acronyms, elected members' names, jargon).

Allow around 20–30 minutes for complete transcription + speaker identification for a 3-hour meeting. AI Studio then generates draft minutes in 1–3 minutes depending on length. Overall, you have draft minutes ready to review less than an hour after the meeting ends.

Yes. The Speakers page (under Vocabulary) accepts Excel imports (template supplied). You can manage several lists in parallel (council, finance committee, CCAS, French CSE…) and each list automatically supports speaker identification, attendance pre-filling and individual vote breakdowns.

Under 30 days for a pilot. SaaS deployment takes one day (organisation creation, sector template configuration, speaker list import), followed by 2–3 weeks of operational support for your first real meetings.

No. Since May 2026, the “Identify from a previous session” feature automatically reuses names from an already annotated session within the same tenant to pre-fill speakers in the new session. Computation takes around 45 seconds on our dedicated GPU infrastructure (France or Europe according to your contractual requirements), then presents a suggestions table to approve in one click. No voiceprint is retained: computation runs in RAM and disappears before the request ends. The feature is disabled by default and enabled per tenant through a dedicated DPA addendum and a data protection impact assessment (DPIA) — we supply the templates.

Retrieve all your data through the portability function (Settings → Portability → Export). The ZIP archive contains all transcripts, minutes, votes, attendance, metadata and source audio files. You can then request complete erasure of your organisation in accordance with the GDPR right to erasure.

Pricing depends on monthly transcription volume, deployment mode (SaaS hosted in France, project-based SecNumCloud, or on-premise) and functional scope. Contact us for a personalised quotation — we adapt the offer to each organisation's context (local authority, practice, clinic, SME).