Can AI Grade a Vintage Movie Poster? A Small Experiment with The Lost World (1960)

finepostervault

Over the past months I have been building a database of my vintage movie poster collection and photographing every piece myself. As many of you know, once you start documenting posters properly you quickly run into the question of condition grading.

Because grading is both technical and subjective, I wondered whether a modern AI system could analyze a poster photo and produce a condition description that would be reasonably acceptable to experienced collectors and dealers.

This post is a small experiment.

---

The Poster

The poster used for the test is:

The Lost World (20th Century Fox, 1960)
US Half Sheet (22 × 28)

I purchased this exact copy at Heritage Auctions, where it was cataloged as Fine+.

For the experiment, I photographed the poster myself and asked ChatGPT to evaluate the condition using a grading style similar to emovieposter.com.

---

My Photography Setup

I wanted a setup that produces a neutral, documentation-style image rather than a “marketing” photo.

Equipment:

Camera: Canon EOS M50
Lighting: 2 × Godox SL60IID (5700K, 60% power)
Light modifiers: 80 × 120 cm softboxes
Color reference: Calibrite ColorChecker Passport Photo 2
Background: Magnetic whiteboard with a non-reflective gray board

Processing:

• Shot in RAW

• White balance only corrected in Adobe Lightroom

• No changes to contrast, saturation, or sharpening

• Exported as JPG

So the image you see is essentially a neutral reproduction of the poster, not an auction-style promotional photo.

Of course I am aware that a full professional evaluation would normally also require:

• photos of the reverse

• detail shots

• possibly raking light

But for this experiment the question is simpler:

Can AI produce a condition description from a single front photo that a knowledgeable collector would consider reasonable?

---

The Photo Used for the Test

(The image shown here.)

---

ChatGPT Condition Assessment (emovieposter-style)

Below is the AI-generated description written in the style typically used by emovieposter.com.

Condition: Very Good

Description:

This half sheet has bright color and a very nice overall appearance. There are pinholes in the corners and several small tears and edge wear along the borders, most noticeable along the upper edge. There are minor corner creases and light handling wear, along with slight waviness near the top border. The image area remains clean and displays well.

---

Comparison with the Heritage Description

Heritage originally cataloged this poster as:

Fine+

Their description mentioned:

• edge wear

• creases and handling wear

• chips and tears along the borders

• surface rippling

Interestingly, the AI result is slightly more conservative, landing at Very Good, which is not dramatically far away from the auction house assessment.

---

Why This Experiment Interested Me

Grading vintage posters is famously difficult because:

• it involves judgment calls

• photographs rarely show everything

• grading standards vary slightly between auction houses

What surprised me is that the AI:

• correctly focused on border damage

• recognized pinholes

• identified edge wear

• and largely ignored the image area, which is relatively clean

In other words, the reasoning followed the same logic most collectors use:
border damage affects grade more than cosmetic wear in blank areas.

---

The Question for the Experts

I would love to hear the opinion of experienced collectors, dealers, and restorers here on VMPF.

Based only on the photograph:

• Is the AI description reasonable?

• Is the grade too harsh, too generous, or about right?

• Do you think AI could become a useful first-pass grading tool for large collections?

Again, I am fully aware that a single front image cannot replace expert inspection. But as a baseline documentation tool, I find the result surprisingly interesting.

Looking forward to your thoughts.

Bruce

It would be "good to very good" for me. Clearly NOT "very good" and clearly NOT "good" so that makes the choice easy.

I WOULD mention that the defects are primarily in the borders, which is very important.

All "good to very goods" are not the same when it comes to displayability.

finepostervault

Thank you.