Updated: March 12, 2026

AI Recognition Quality

Pain Tracker can fill diary entries from short text messages and Telegram voice notes. Text is processed by a lightweight recognition model. Voice notes are transcribed first, then passed through the same extraction flow.

How It Works

  • Text messages go directly to the recognition model.
  • Telegram voice messages are transcribed first, then parsed into fields such as pain level, triggers, medications, symptoms, and pain location.
  • The recognized data is used to fill the log automatically.

Examples

  • Text: "Yesterday 6/10 on the left, stress, took ibuprofen, helped."
  • Voice note: "Pain level 6 on the right after office took sumatriptan."

Supported Languages

  • Best supported end-to-end languages: English, German, and Russian.
  • Set your app language to the same language you write or speak.
  • Other languages may partially transcribe, but they are not currently supported in the full recognition flow.

Current Limitations

  • Voice recognition currently works with Telegram voice messages.
  • Voice messages longer than 30 seconds or larger than 5 MB are rejected.
  • Background noise, multiple speakers, very fast speech, mixed languages, or long unstructured stories can reduce accuracy.
  • AI can miss details or put them into the wrong field.

Best Results

Keep messages short and structured.