TranscrIA
Long meeting recordings become corrected, speaker-labelled transcripts, structured summaries, and Word minutes in TranscrIA, with no per-minute bill. An audio check scores each upload and plots a difficulty timeline before transcription starts, and a single slider then picks the deliverable, from plain subtitles to a full quality package. Detected speakers arrive with playable excerpts, talk time, and an acoustic gender hint, and you confirm names, participants, and a domain glossary, while attached PDF, Word, or PowerPoint files ground the summary in the real agenda. A local OpenAI-compatible model corrects the text, and a waveform editor with speaker lanes handles splitting, merging, undo, and restorable versions. Chat lets you question a finished transcript and apply a fix across every deliverable, or ask across all meetings you may see and get cited answers plus a My Actions list. Eighteen meeting types shape the Word minutes, and meetings on Jitsi, Zoom, Visio, and Google Meet can be captured live over outbound-only connections. Roles, groups, single sign-on through OIDC or LDAP, personal API tokens, and an exportable audit trail suit shared use. Transcription runs on NVIDIA GPUs or, with the Kroko engine, on CPU alone, and an approved cloud LLM can write summaries when no GPU is present. The interface speaks French, English, Portuguese, and beta German, Spanish, and Italian. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.
Deploy