01
Background processing
After upload, a queue extracts text, builds search embeddings, and analyzes colors — so the gallery feels fast while enrichment catches up quietly.
psychology
Practical AI
AI that earns its place
We do not sprinkle AI on every button. prn.tf uses vision and embeddings where they actually help — finding text in screenshots, suggesting tags, and making your library searchable without manual busywork.
Philosophy
Useful, not performative
Heavy work runs in the background. Interactive AI stays in the gallery, on your terms.
02
On-demand in the UI
OCR and auto-tagging run when you ask for them from a photo — not on every pixel of every upload by default.
03
Search that respects content
Extracted text and embeddings power lookup by what is inside the screenshot, not just the filename you forgot to rename.
Capabilities
What runs under the hood
Vision OCR
Readable text is pulled from UI shots, terminals, docs, and chat windows. Run it from the photo panel or let the upload queue handle it automatically.
Auto-tagging
Vision suggests short, practical tags — app names, content type, language hints — so collections stay tidy without typing every label yourself.
Text embeddings
OCR output is embedded for semantic search across your library. Find the error message or UI string you remember, not the upload date.
Image captions
One-line descriptions summarize what a shot shows — useful for skimming large galleries and understanding context at a glance.
Workflow
From upload to findable
- check_circle Upload via client → shot appears in gallery immediately
- check_circle Queue enriches OCR text and embeddings when configured
- check_circle Open a photo → extract text or suggest tags on demand
- check_circle Search and browse with tags, text, and color metadata together
upload
instant
ocr queue
background
auto-tag
on demand
search
by content
See it in your gallery
Upload a few screenshots and try OCR or tag suggestions on a real shot.