Photos, PDFs and voice messages
The AI also interprets media in the conversation, not only text. For how it works in general, see AI overview; for what is structured into vehicles and parts, see Recognition.
Photos
Images are read and the data enter the conversation as if they were text. Typical recognized content:
- plates
- registration booklets
- labels and packaging
- catalog or diagnosis screenshots
- documents
- photographed components
Photo and caption are treated as one message.
If a photo is still uploading, the AI waits until it is ready before processing later messages.
PDFs
PDFs (quotes, invoices, delivery notes, parts lists, scanned booklets) are read page by page:
- one line per spare part
- vehicle with plate and VIN when present
- without prices and without personal data
Voice messages
Voice messages are transcribed and interpreted as text. Plates and codes spoken aloud are reconstructed.
Cautious reading
Reading is more careful than before:
- partial plates
- VINs inconsistent with the make
are not accepted silently: they are flagged as “Uncertain reading”.
“AI readings” in the viewer
Opening a photo or PDF full screen shows the “AI readings” panel.
It shows:
- what the AI read in that file
- what the request currently contains
- the pencil to correct plate, VIN or code from there
- uncertain readings
- how many PDF pages were read
- any unread files
Correction from here follows the same rules described in AI panel and corrections.
The toggles for voice transcription, image reading and text recognition are in Settings → Artificial intelligence → AI recognition (see also Display modes).