Fidelium.
by Rubber Raccoon
EST. 2026

Old photo restorationRestore the photo.
Keep the person.

Faithful to every face.
Old photo restoration for damaged family photos. Every face is rebuilt from that person’s own face in this photo — or from other photos of them you add — and checked against your photo, to keep it the same person. Scratches, fading and torn corners are repaired, and a written record tells you what happened to every face.
NO SUBSCRIPTION · PAY PER PHOTOGRAPH
+
DROP A SCAN OR PHOTO OF A PRINT
JPG · PNG · TIFF — USED FOR THIS RESTORATION ONLY
ARCHIVE Nº 21 · MARCH 19944 FACES
Before: a faded, scratched 1994 kitchen birthday photo of a family of four After: the same birthday photo, restored
01
02
03
04
BEFOREAFTER
FACE 01REBUILT · FROM THIS PHOTO
FACE 02REBUILT · FROM THIS PHOTO
OTHER 2 FACESREBUILT · FROM THIS PHOTO
SAMPLE PHOTO GENERATED WITH AI · RESTORED BY FIDELIUM
 KEPT AS IT WAS · ENLARGED  ·  REBUILT · REF 04  ·  PRESERVED — TOO SMALL TO REBUILD  ·  CHECK PASSED  ·  RUN 2 REJECTED — NO CHARGE  ·  KEPT AS IT WAS · ENLARGED  ·  REBUILT · REF 02  ·  CHECK PASSED  ·  PRESERVED — NO CHARGE  ·  KEPT AS IT WAS · ENLARGED

THE ZOOM TEST

Zoom in. It’s still him.

Before: the faded 1994 birthday photo
After: the same photo restored, the grandfather rebuilt from his own face
01020304
BEFOREAFTER FACE 04 · REBUILT · FROM THIS PHOTO
  1. 01The photo you have. Faded, scratched — four faces.
  2. 02Zoom into one face. This blur is all the photo still knows about him.
  3. 03Rebuilt from his own face in this photo. Drag the line and compare for yourself.

SAMPLE PHOTO GENERATED WITH AI · RESTORED BY FIDELIUM

THE ARCHIVE

Every decade. Every face.

Nº 01 · 1946 · WEDDING2 FACES
Before: a 1946 wedding portrait, torn and creased After: a 1946 wedding portrait, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 01 beforeFace 01 after
01 · REBUILT
Face 02 beforeFace 02 after
02 · REBUILT
TORN · CREASEDREBUILT · FROM THIS PHOTO
Nº 02 · 1955 · REUNION11 FACES
Before: a 1955 family reunion of eleven people, creased and faded After: a 1955 family reunion of eleven people, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 05 beforeFace 05 after
05 · REBUILT
Face 01 beforeFace 01 after
01 · REBUILT
Face 04 beforeFace 04 after
04 · REBUILT
Face 02 beforeFace 02 after
02 · REBUILT

+ 7 MORE FACES IN THIS PHOTO

CREASED · FADEDREBUILT · FROM THIS PHOTO
Nº 03 · 1962 · SUMMER3 FACES
Before: a 1962 lakeside family snapshot, faded and scratched After: a 1962 lakeside family snapshot, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 01 beforeFace 01 after
01 · KEPT
Face 02 beforeFace 02 after
02 · REBUILT
Face 03 beforeFace 03 after
03 · PRESERVED
FADED · SCRATCHEDKEPT · REBUILT · PRESERVED
Nº 04 · 1994 · BIRTHDAY4 FACES
Before: a 1994 kitchen birthday party, faded and scratched After: a 1994 kitchen birthday party, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 01 beforeFace 01 after
01 · REBUILT
Face 02 beforeFace 02 after
02 · REBUILT
Face 03 beforeFace 03 after
03 · REBUILT
Face 04 beforeFace 04 after
04 · REBUILT
FADED · SCRATCHEDREBUILT · FROM THIS PHOTO
Nº 05 · 1997 · BEACH4 FACES
Before: a 1997 beach holiday photo of a family of four, faded and dust After: a 1997 beach holiday photo of a family of four, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 01 beforeFace 01 after
01 · REBUILT
Face 02 beforeFace 02 after
02 · REBUILT
Face 03 beforeFace 03 after
03 · REBUILT
Face 04 beforeFace 04 after
04 · REBUILT
FADED · DUSTREBUILT · FROM THIS PHOTO
Nº 06 · 1999 · CHRISTMAS5 FACES
Before: a 1999 Christmas Eve living-room photo, faded and dust After: a 1999 Christmas Eve living-room photo, restored BEFOREAFTER

FACE BY FACE · BEFORE | AFTER · CLICK TO ENLARGE

Face 01 beforeFace 01 after
01 · REBUILT
Face 02 beforeFace 02 after
02 · REBUILT
Face 03 beforeFace 03 after
03 · REBUILT
Face 04 beforeFace 04 after
04 · REBUILT

+ 1 MORE FACE IN THIS PHOTO

FADED · DUSTREBUILT · FROM THIS PHOTO

SAMPLE PHOTOS GENERATED WITH AI · EVERY "AFTER" IS A REAL, UNEDITED FIDELIUM RESULT · DRAG ANY PHOTO TO COMPARE · CLICK IT TO ENLARGE

HOW IT WORKS

Three steps, and they’re back.

A creased, faded 1955 family reunion photo with its eleven faces marked010203040506070809101111 FACES FOUND
01
We measure first.
We find every face and measure how much the photograph can still prove. Where the evidence runs out, we do not invent a face — it is preserved and marked, at no charge.
Face 05 before, grainy with a scratch across it, and after, rebuilt from his own face in the same photoFACE 05 · BEFOREAFTER
02
The person comes from the person.
Each damaged face is rebuilt from that person’s own face in this photo. Have other pictures of them? Add 3–8, any decade, with their age in this photo — the rebuild is made from them, not from anyone else’s face.
The 1955 reunion photo restored, all eleven faces rebuilt and checkedCHECK PASSED · 11 OF 11 FACES
03
Checked before you see it.
Every rebuilt face is checked automatically against your original. A run that fails the check is redone — and a rejected run never costs you anything.

YOUR OTHER PHOTOS

A tiny face? Add their other photos.

The smaller and blurrier a face, the more any AI has to guess. Add 3–8 other photos of that person — from any decade, with their age in this photo — and the rebuild is made from them, not from anyone else’s face. Fidelium suggests it when a face has little detail.

A father's face in a faded, scratched 1994 camping photo, only a few millimetres wide in the print
IN THE 1994 PHOTO
The same man around 1988The same man around 1999The same man around 2005
HIS OTHER PHOTOS · 1988 · 1999 · 2005
Rebuilt from the 1994 photo alone
REBUILT FROM THIS PHOTO ONLY
Rebuilt with his three other photos; the hat and shirt are the 1994 photo’s own
REBUILT WITH HIS PHOTOS

SAMPLE PHOTOS GENERATED WITH AI · BOTH RESULTS ARE REAL FIDELIUM RESTORATIONS

EVERY FACE IS LABELED

Every face labeled. Nothing invented.

KEPT AS IT WAS · ENLARGEDThe face was clear enough to keep. It keeps its own features — cleaned and enlarged with the rest of the photograph, not rebuilt.
REBUILT · FROM THEIR OWN FACEThe face was rebuilt — from that person’s own face in this photo or from your photos of them, checked, and marked. You can run it again, mix runs, or go back to the real pixels at any time — free.
PRESERVED — TOO SMALLWhen there isn’t enough evidence for an honest rebuild, we don’t sell you one. The face stays as your photograph shows it, softly enlarged — at no charge.

THE RULES WE DO NOT BREAK

The promises that make it safe to hope.

Never.
  • Never rebuilt from a stranger’s face — every rebuild starts from that same person’s own face.
  • Never invention sold as restoration — anything rebuilt or filled in is labeled in the record.
  • Never a subscription for a one-time job — you pay per photograph, once.
  • Never a charge for a run that our own check rejected.
Always.
  • Your original is always kept — one click returns any face to its real pixels.
  • A written record with every restoration: what was kept, what was rebuilt, what was preserved.
  • Your references are used for this restoration only, then deleted.
  • You pick the final face — run again, compare runs, mix freely.

QUESTIONS, ANSWERED STRAIGHT

Questions people actually ask.

Can AI restore an old photo without changing the person’s face?
Yes — that is what Fidelium is built for. Many AI enhancers repaint faces from their training data, which is why restored relatives can look like strangers. Fidelium rebuilds each damaged face from that person’s own face in the photo, or from other photos of them that you add, checks every result automatically against your original, and labels every face in a written record: kept, rebuilt or preserved.
What do I need to restore a damaged family photo?
A scan or a phone photo of the print — that is enough, because each damaged face is rebuilt from the photo itself. For a badly damaged face, adding 3–8 other photos of that person (any decade, with their age in this photo) helps the rebuild stay true.
Can Fidelium repair torn, creased or scratched photos?
Yes. Scratches, dust and creases are removed automatically. For long scratches and deep creases there is a deeper clean (one more credit), which can also be added after the restoration for two credits. Torn-off corners are filled in and marked as filled.
Do you colorize black-and-white photos?
No. Black-and-white and sepia photos stay black-and-white — Fidelium restores, it does not recolour. Faded colour photos get their colours back.
How large can I print a restored photo?
Every photo is enlarged to up to 24 megapixels — enough for a print of about 50 × 34 cm. An extra-large file for bigger prints is two more credits.
What happens to faces that are too small or too damaged?
When a face is too small for an honest rebuild, Fidelium does not invent one: it stays as the photo shows it, gently enlarged, marked in the record and never charged. People who are out of focus in the background stay soft, as the camera saw them.
How many faces can one photo have?
Up to 60. Two credits cover a photo with up to 10 faces; a larger group takes two more credits per started 10 faces — a class photo with 29 children is 6 credits.
Can I use other photos of my grandmother to rebuild her face?
Yes. Add 3–8 photos of her from any decade and type her age in this photo. The rebuild takes her face from them but keeps the age, hair and expression of this photograph.
Can Fidelium restore pets too?
Yes, on request: switch on “find pets” and dogs, cats and horses are found and rebuilt like faces.
What is the difference between FAITHFUL and CRISP?
Two modes for rebuilding a face. FAITHFUL is the standard and stays closest to the real person. CRISP draws bolder detail but can make people look older or more polished; it is two more credits per face.
How much does photo restoration cost?
Fidelium sells credits, never subscriptions: 30 credits $9, 80 credits $19, 250 credits $49 — about $0.39 to $0.60 per photo. Two credits restore a whole photo with up to 10 faces. Re-running faces is 1 credit per 5 faces; removing every scratch is one more credit (two if added afterwards); an extra-large file and CRISP (per face) are two more each. The credits sit in one Rubber Raccoon wallet that also works in our other tools; new accounts get 3 free credits — enough to restore your first photo in full before you buy.
How long does a restoration take?
Usually one to two minutes per photo. Large group photos and the deeper scratch clean take a little longer.
Is the likeness guaranteed?
No restorer can promise that — a photo may not hold enough of a face. You see every face before you keep it, and can run any face again, compare runs, or keep the original face for free. Runs you do not like are not refunded; runs rejected by our own check are never charged.
Do you use my photos to train AI?
No. We use your photographs and reference photos only for your restoration and never to train AI models. They are deleted 30 days after the run. Details in the privacy policy.
One photograph. One promise.
2 CREDITS PER PHOTOGRAPH · FROM $0.39 RE-RUN 1 CREDIT PER 5 FACES · REJECTED RUNS FREE
PHOTOGRAPH
Your photograph
FACE REGISTER
Rebuilt faces come from your reference photos. Nothing is invented; where the evidence runs out, the face is preserved and marked.

FAITHFUL is our standard: it stays closest to the real person. CRISP renders bolder detail but can make people look older or more polished — try it on any face if you like.
Restoring…
This usually takes one to three minutes. Every rebuilt face is checked before it reaches you.
RESULT — BEFORE / AFTER ↓ DOWNLOAD
BeforeBEFORE
AfterAFTER
RESTORATION RECORD
Not quite them? Run the face again — each run is a fresh roll. Some faces take a few runs, sometimes on the other engine, and a perfect likeness can't be guaranteed. You can always pick ORIGINAL in a face's runs to keep it as it was.
FACE 01
Show us this person in 3–8 other photos. Any decade works — clear face, no sunglasses. We rebuild only what they can vouch for.
Are they younger or older here than in your references? Tell us roughly how old — we keep the age this photo shows, not the references'.
Only use photos of people who have agreed — or yourself. We use your references only for this restoration, never to train AI models, and delete them after 30 days.