What happens on this website today
The revised public page does not request microphone access. Its interactive demonstration transforms text only. It does not record, transcribe, upload, or retain spoken audio.
Current boundary: The public website is a concept and interface demonstration. It is not the live Kinder Voice engine.
Rules for any future voice product
- Microphone access must be explicit and visible.
- The active processing mode must be understandable in plain language.
- The user must be able to bypass the transformation immediately.
- Any processing outside the user’s device must be disclosed before use.
- Privacy claims must be verified against the actual packaged product, not inferred from the design intent.
Retention should default to none
Raw audio should not be retained by default. Diagnostics should avoid voice content whenever possible. Any optional recording, model-improvement contribution, or troubleshooting capture must require a separate, informed action.
Network behavior must be inspectable
If future versions use hosted inference, relays, telemetry, crash reporting, or account services, the product should identify those services, the data sent, the reason, and the retention period. “Local-first” must describe measured behavior, not aspiration.