JPG, PNG, WEBP. One photo or several (max 8, 24MB each).
Luna Lisa Image Model
Community stills from X testers. Not a client list. Each frame is credited below.
What is the Luna Lisa image model?
Luna Lisa is the image model this page is for. Testers on Arena and X spent August 2026 stacking it against GPT Image 2: phone-photo skin instead of plastic shine, poster type you can actually read, and a newer knowledge cutoff. On lunalisa.cc you generate in the browser. Luna Lisa is the VIP model on the panel. Guest trial stays on Spark, text-to-image only.
OpenAI’s published API name is still GPT Image 2. luna-lisa-alpha is the Arena catalog label testers used. This site is independent. We are not OpenAI. The stills below are community tests, credited to the posters — not a promise that your next run copies that frame.


What testers say is new in Luna Lisa
Three claims show up again and again. None of them is an official spec sheet.
- Less plastic, more phone photo. Oil-shine skin, night-flash group shots, freckles, wet leaves. That is the most consistent praise for Luna Lisa and for Mona Lisa 1 before it.
- On-image text. chetaslua called out great text rendering. mirochill showed Japanese and English that you can read — and warned Luna Lisa still loves dumping extra words into a scene.
- Newer knowledge. Mona Lisa 1 testers got 2025 dates and a living Charlie Kirk. Luna Lisa testers posted the opposite headline. Heuristic, not a benchmark set.
Weak spots they also posted: hands that should be holding each other, dirty grain on clouds and leaves, and whether this is a generation leap or a small step. synthwavedd spent a day on Mona Lisa 1 and said slightly better, not a jump.



More Luna Lisa stills
Tester frames, shown as-is. Credit stays with the posters. Not a promise that your next generate copies that still.




Stills are community Arena / X tests. Not generated on this form.
What testers posted on X about Luna Lisa
Original posts. If a widget fails to load, the quote and link stay.
New GPT-image is testing. luna-lisa-alpha is being tested. this is new checkpoint last one was monalisa. recent knowledge cutoff. great text rendering. super realistic.
— chetaslua (@chetaslua) August 19, 2026
Better than mona. Huge jump in character consistency. Grain seems gone. Still loves generating lots of text in the image.
— mirochill (@mirochill) August 19, 2026
luna-lisa-alpha vs GPT Image 2.0 vs Nano Banana Pro.
— Harshith (@HarshithLucky3) August 19, 2026
mona-lisa-1. prompt: make a image of your creator.
— chetaslua (@chetaslua) August 9, 2026
Tested all day. Slightly better realism and noise than GPT Image 2. Not a generational leap. The “-1” means another checkpoint is coming.
— synthwavedd (@synthwavedd) August 9, 2026
OpenAI Verify shows SynthID. C2PA is often missing.
— Token Gremlin (@TokenGremlin) August 9, 2026
Convenience-store flash comparison. Plastic look is down.
— WolfRiccardo (@WolfRiccardo) August 9, 2026
Complex backyard scene. GPT Image 2 could not hold the groups and the signage.
— WolfRiccardo (@WolfRiccardo) August 10, 2026
How to generate with Luna Lisa on this site
Three steps. The tool is the prompt box at the top of this page. Chrome, Safari, Edge, or Firefox. No install.
Open lunalisa.cc
The generate path is this page. No app. No sign-up to start the guest trial.
Start on Spark
The panel opens on Spark, our free text-to-image model. Switch to Luna Lisa when you want the VIP path. Photo edit needs Luna Lisa.
Prompt, generate, check
Name the subject, the light, and any text you want in the frame. Check faces, hands, and labels before you download.