Bring the fragments.
Choose your photographs or explore one of four demo archives. Johanna, Mira, Inna and Ron are fictional names for curated selections of licensed everyday photographs.
IMAGE → LOCAL PREVIEW01 / WHERE THE THREAD BEGANINSPIRED BY THE ORIGINAL THESEUS
FROM FRAGMENTS TO REFERENCES
A collection becomes useful when its fragments can be traced back to their sources.
Licensed demo photographs illustrating the Theseus idea. These are not the original creator’s photographs.
THE ORIGINAL SYSTEM / CHOLY (@CRYPTOCHOLY)
Before the interface,
there was
a personal archive.
Theseus began with choly (@cryptocholy), who asked what years of photographed art could reveal about personal taste. The creator reports turning 19,651 phone photos into 1,541 identified artworks across 63 venues and 19 countries, working locally on a Mac.
The result was a map and a portable reference file for creative work. This browser project follows that idea into everyday photographs.
These are author-reported figures for the original Theseus project, not this prototype's dataset or independently verified results.
INSPIRED BY CHOLY (@CRYPTOCHOLY). INDEPENDENTLY DEVELOPED.02 / THE IDEAATTENTION → MEMORY → CONTEXT
You already have
a language.
It is in your archive.
LOCAL DETECTIONS / SOURCE FRAMES
A photograph becomes a set of visible details. The frame remains the reference.
Illustration from the licensed demo archive. Boxes come from local object detection; they do not identify people.
A family table. A day with friends. A small moment you wanted to keep.
Your photographs hold a record of what catches your attention. Theseus turns those fragments into an archive you can explore, label and carry into your next idea.
For artists, designers, curious people — and anyone whose camera roll deserves a second look.
FREE CORE / NO ACCOUNT OR WALLET REQUIRED03 / FOLLOW THE THREADFROM A FRAGMENT TO A REFERENCE
Keep what moved you.
See what connects it.
A reference library shaped by your own attention. Start small; give each image a little context.
Give them context.
Colors and face/object regions are processed locally. The boxes are model detections, not identities. Add context yourself and review any suggested labels.
LOCAL REGIONS + YOUR LABELSFollow a connection.
Switch archives on the left; read recurring details, spaces and gestures on the right. Each demo reading is editorial, with supporting frames you can inspect. Optional cloud reading of your own archive requires explicit consent.
MAP → PORTABLE CONTEXTA CONNECTION, UP CLOSE / ILLUSTRATIVE EXAMPLE
One detail.
Another way in.
A repeated color can connect photographs that otherwise have little in common. It is a visual motif, ready for your interpretation.
REFERENCE FIELD / 003 FRAGMENTSLABELS REVEAL A PATH. YOU DECIDE WHAT IT MEANS.
UNDER THE SURFACE / HOW THE READING WORKS
First the details.
Then the story.
The eye notices a detail before the mind finds a story. The archive follows that same order: first a careful look at each photograph, then a search for what returns.
ILLUSTRATED SEQUENCE / CACHED LOCAL DETECTIONS
Scene-level threads compare detected group compositions. They do not identify people or establish a shared event.
Detection is not identification. The system does not establish who people are, who is related, or who met whom.
Prepare on your device.
Your photographs begin here, on your device. The browser makes smaller previews, leaves the original files untouched and removes their embedded EXIF metadata from these copies. Local tools can notice colors, objects, faces and written words; you decide which labels to keep.
ON DEVICE / MEDIAPIPE · MOBILECLIP · TESSERACTRead each photograph.
If you invite a cloud reading, selected previews and their labels go to the service you approve: Google Gemini or Venice. Up to 100 photographs are read in small groups. Each keeps its own observation, so a later connection can always lead you back to the frame.
OPTIONAL CLOUD / IMAGE → OBSERVATIONCompare the observations.
Then the observations are read together. A familiar room, a gesture, the same kind of gathering: details may return across photographs taken far apart. This second reading uses the descriptions already made, with a source frame behind every proposed connection.
ARCHIVE READING / OBSERVATIONS → CONNECTIONSReturn to the evidence.
A connection is an invitation to look again. Open its photographs, compare what is visible with what is suggested, and decide whether it rings true. Unread frames remain marked as unread; an unfinished reading keeps its gaps.
YOUR REVIEW / FINDING → SOURCE FRAMES
Available now: four demo archives with editorial readings, local face/object regions, visual categories, label OCR, map and export. Save your own private cloud album after signing in. Cloud archive reading and artwork suggestions require separate consent.
04 / YOUR CONTEXTSMALL ARCHIVE. SPECIFIC POINT OF VIEW.
Give the next idea
somewhere to begin.
One selection, three ways to carry it.Demo / 03 references
A brief with a memory.
Gather the textures, colors and references behind a project. Export a Markdown palette to use in a creative brief or share with an AI tool of your choice.
An archive you can carry.
Save compressed previews in this browser, or sign in to save a private album on this site. Export labels as JSON or Markdown and keep a backup.
A deliberate choice to share.
Your photos stay on your device until you choose cloud storage or consent to online analysis. Cloud albums start private; publishing is a separate choice. AI readings saved with an owned album remain private even when the album is published.
05 / MEMORY STUDIOYOUR ARCHIVE, MADE INTO SOMETHING
MEMORY STUDIO / FREE TO EXPLORE
Keep the photographs.
Make the story.
A photograph keeps a moment still. Memory Studio gives it another form: a film with a beginning and an ending, a chapter to read, or a question worth following.
Bring your own photographs and tell us what you want to make. You see the plan and the full cost before choosing to produce it.
Let several moments become a film.
Choose 2–6 of your photographs and describe the story. AI proposes an arc for you to review, then animates each approved frame into a scene. You receive one film with sound: 8–24 seconds, in a single MP4. A simple photographic montage stays free on your device.
Choose a mood, let the story suggest one, or bring your own music.
YOUR PHOTOGRAPHS + YOUR DIRECTION → ONE FILM
Chapters with something behind them.
Build a chapter from each supported finding. Every observation keeps its source photographs beside it. Download a standalone illustrated story to keep or share deliberately; creating it does not publish your album.
EXISTING READING → ILLUSTRATED STORY
Follow one question further.
Ask what repeats across your journeys, rooms or celebrations. With your consent, your selected service, Google Gemini or Venice, compares the observations from an existing reading, without uploading the photographs again. Answers link back to source frames and can say the evidence is insufficient.
OPTIONAL CLOUD / BOUNDED REQUEST
A new chapter, not a new beginning.
An archive grows by addition. Previously read photographs keep their observations; only new or changed images need another visual reading. The whole story can then be reconsidered. Open the same unchanged archive again and no AI request is made.
SAVED OBSERVATIONS → NEW FRAMES
PROJECT-SUPPORTED CREATION
Make something personal.
Know the cost first.
A story plan costs the project no more than $0.02. Film production depends on the chosen service and 2–6 scenes; a fresh quote shows the complete cost before support. Neither uploading photographs nor reviewing a plan starts video generation.
For an eligible production, the project covers 30% when its production budget is available. Your quote shows the full cost, that contribution and the remainder payable in the project token. You approve it before production begins.
Explore and make local previews freely. A wallet is needed only for token checkout. The Solana network fee is separate and paid in SOL.
FROM YOUR ARCHIVE TO AN OUTPUT01 → 04
- 01SOURCE ARCHIVE
Begin with the photographs and observations you already have.
- 02CHOOSE OUTPUT
A generated film, a photo montage, an illustrated story, a focused question or an archive update.
- 03SEE YOUR QUOTE
Output scope · project contribution · amount payable
- 04CONFIRM AND CREATE
Review the quote before any token payment. A paid job starts only after payment is verified and compute is reserved.
- LOCAL STUDIO PREVIEW
- OPEN / NO WALLET
- PROJECT TOKEN PAYMENTS
- QUOTED PER PRODUCTION
HOW A FILM IS PAID FOR
From a token,
to a finished memory.
Your payment has a purpose: to cover the services that turn your photographs into a film. The project contribution reduces your share; the remaining amount is quoted in the project token.
The token amount can change with the market. Each quote uses a current SOL/USD rate and a Jupiter market quote, then fixes the amount for that checkout. A separate SOL network fee is shown by your wallet.
A PUBLIC PATH FOR PRODUCTION PAYMENTS
Verified token payments go to the public payment wallet below. The payment service sells received project tokens at market to fund production costs. These sales create sell pressure; they are not a token burn or a reward to holders.
AHxrYbt4nPErECQceaFTVhP9EWNys8CpnLqVf8YMdgdAPayment receipts are visible on Solana. Sales start only after a verified payment and an executable route.
Tokens received and provider credit are separate balances. A sale alone does not pay the AI provider; production also needs funded provider credit and a reserved budget. If those are unavailable, a new production does not start.
Solana · SOL pair · standard Pump.fun fees. Each production quote shows the project contribution and your token payment.
06 / CURATORSFREE TO EXPLORE / NO WALLET
A PUBLIC STUDY OF ORDINARY LIFE
One frame becomes
a wider view.
Explore 140 licensed everyday photographs, then follow public chapters shared by their owners. Branches group source frames by collection and supplied dates; similar faces never become claims about identity.
Your album joins Curators only when you choose to include it. New photographs added to that public album extend its branch. Make it private or remove it from Curators to take it out of the shared tree.
07 / BEFORE YOU STARTA FEW THINGS WORTH KNOWING
Keep the curiosity.
Know the details.
What does the prototype do today?
Explore four clearly marked demo archives or add your own photographs. Local models find face/object regions, suggest visual categories and transcribe label text for review. Explore editorial demo readings and follow each observation to its supporting frames. Optional Gemini analysis of your own photos requires explicit cloud consent.
Can I use it on my phone?
The interface adapts to desktop, Android and iPhone screens. Real-device checks are still pending. JPEG, PNG and WebP are the safest starting point; HEIC support depends on your browser.
Where do my photos go?
Local tools read your selected photos in the browser. Save on this device stores compressed previews and labels in that browser. Saving a cloud album requires sign-in and explicit upload consent; compressed copies are stored privately on this site until you choose to publish. Whole-archive AI reading separately asks permission to send compressed photographs and catalog labels to the service you approve, Google Gemini or Venice. The demo combines historical film and contemporary color photographs, hosted here with credits and licenses linked below.
What travels in an export?
JSON and Markdown contain text references and color labels, not the photographs. Capture dates, GPS and original file paths are excluded. Review any names or places you have added before sharing.
Do I need a wallet or a token?
No. The core archive and local tools are free. A Solana wallet is used only when you choose token checkout for an optional production. The quote shows any project support before you confirm.
How will AI read a whole archive?
With your explicit consent, Gemini 3.8 Flash processes available photographs in groups, then connects successful observations into one reading. This version accepts up to 100 photos and 100 references, with supporting frames and an honest coverage report. Missing or unreadable photographs are listed. Shared pilot compute allowances apply; the service checks its remaining budget before starting a paid request. People sharing an IP may share a rate limit. Demo readings are curated by a person, not generated by AI.
Is this an AI personality test?
It is a tool for exploring references you chose. Editorial readings describe visible details in the demo photographs. Local boxes are model detections; optional AI readings are interpretations with evidence. Neither establishes identities, family ties or meetings. Their meaning is yours to interpret.
08 / KEEP IN TOUCHTHE NEXT CHAPTER IS TAKING SHAPE