Provenance

Know where every image came from.

Camera, AI or drawing. Filamama will keep the prompt, the source images, every edit and C2PA Content Credentials, on your own server.

In development. Not in a release yet.

192.168.1.20:5400
Type
AI-generated
Prompt
“Our lighthouse from the Brittany trip, at sunset, in watercolour”
Model
Image model · seed 4127 · 30 steps
Sources
2 photos
Credentials
C2PA · signature valid

Edit history

  1. Generated12 Aug, 21:04
  2. Cropped to 4:512 Aug, 21:06
  3. Warmer colours12 Aug, 21:07
A mock-up of the Origin tab, in development: type, prompt, model, source photos, Content Credentials and edit history.

A mock-up of the Origin tab we are building. Details will change.

The plan

Four things, kept with every image.

Photos and AI images now sit side by side in the same library. Years later, you should still be able to tell them apart.

1

Camera, AI or drawing.

Each image gets a type: taken with a camera, generated by AI, or drawn.

  • Set from the camera details in the file or its Content Credentials when they say so
  • You can set or correct it yourself
  • Filter the library by type

2

What made an AI image.

For AI images, Filamama keeps what it took to make them.

  • The prompt
  • The model and its settings
  • The source images, linked to the originals in your library

3

C2PA Content Credentials.

Many cameras, phones and AI tools sign their images with Content Credentials. Filamama will read them and keep them.

  • Show what the credentials say, and whether the signature checks out
  • Keep them with the original, unchanged
  • Leave them in when you download or export

4

Every edit, in order.

An edit history for each image: every step from the first version to the last, like a timeline.

  • Each step with its date and what changed
  • The earlier versions stay available
  • Edits made in Filamama and, when the file says so, before it arrived

For AI assistants

Ask where an image came from.

Filamama's MCP server will expose provenance too. An AI assistant with a token for your library can answer questions like these.

  • “What prompt made this picture?”
  • “Show me the originals this was made from.”
  • “Which images in the Brittany folder are AI-generated?”
  • “Has this photo been edited?”

Same rules as today: read-only by default, one account per token, every call logged.

Assistant
> What prompt made this picture?

Generated on 12 Aug from 2 photos in
Trips. Prompt: "Our lighthouse from
the Brittany trip, at sunset, in
watercolour". Then cropped and warmed.

Illustration. Not a real conversation.

Questions.

Why does this matter in a family library?

In twenty years, someone will scroll back to a picture of their grandmother and wonder if it really happened. AI images already look like photos. If the library writes down where each one came from when it arrives, the answer is still there later.

What is C2PA, in plain words?

C2PA is an open standard for Content Credentials: a signed label inside an image file that says which camera or app made it and how it was edited. Some cameras, phones and AI tools already add one. Because it is signed, a change to the image or the label shows up when the signature is checked.

Does Filamama do this today?

No. Provenance is in development and not in any release yet. This page describes what we are building.

Can Filamama tell if a photo is AI without a label?

No, and it won't guess. It records what it knows: the camera details in the file, the Content Credentials if there are any, and what you or the tool that made the image tell it.

See what Filamama does today