Recordings of conversations deserve careful handling
Below is what we can state today. Items marked as placeholders are decisions for the web version that the owner still has to confirm — we would rather leave them visible than claim something we cannot support.
Voice profiles are biometric data
Voice recognition is the one feature that processes something as personal as a voice print. It is treated accordingly.
Voice prints are stored encrypted and are never included in transcripts, exports, subtitles or AI documents. Only names appear there.
Deleting a voice profile deletes all of its voice samples. Individual samples can be removed from a profile at any time.
Voice recognition can be switched off for the account or for a single file. With it off, speakers are still separated — they just stay unnamed.
On noisy or phone recordings recognition lowers its own confidence and asks for confirmation instead of naming someone automatically.
The app will not save a voice profile without a recorded legal basis or consent. The field is required, stored with the profile, and visible in the profile list.
What AI does and does not do
Each decision, task and objection carries the timestamp of the sentence it came from, so any reader can check it in one click.
Items with no basis in the transcript, invented names and invented deadlines are dropped, and the document reports how many were removed.
When you edit the transcript after generating a document or subtitles, they are flagged as outdated rather than silently left wrong.
AI output is a draft you can edit as markdown, regenerate, or reject. Roles, speaker names and scorecard criteria stay under your control.
Open for the web version
These lines will be replaced with commitments once the owner decides. Until then they stay marked.
Accessibility
The app targets WCAG AA contrast, visible focus states on every control, and full keyboard operation of the editor. Shortcuts are shown in tooltips so they can be discovered without documentation.