GPT Image 1.5｜True-Photo Edits
Describe it or upload a photo — only what you ask changes, the rest stays pixel-perfect. One selfie is enough.
GPT Image 1.5-style generator and editor: text-to-image true-photo portraits plus targeted edits that keep the rest pixel-perfect. Try it free on Loomi today.
Features
- Targeted edits——Hair color, clothes, background, a dog — only the part you name changes. Identity, light and composition stay pixel-identical, before and after stack up cleanly.
- True-photo finish——Pores, flyaway hair, grain and slight overexposure are kept on purpose. No plastic airbrushed skin; finger counts and teeth get self-checked so posts survive scrutiny.
- Progressive refine——Not happy with round one? Keep editing: stance first, then atmosphere, one variable at a time. It gets steadier round by round instead of mushier.
- Clean poster text——Sale posters and event teasers in short English, clean typography, every letter spell-checked before delivery. Goodbye garbled signboards.
How it works
- Say What You Want Upload a well-lit front-facing selfie, pick the edit-upload mode, and spell it out: bright silver-white hair, everything else unchanged.
- Generate in Seconds Hit generate and get one image in tens of seconds: face, clothes and background kept, only the hair color changed.
- Refine Until It's Right Change one variable per round; roll back if it mushes. Download when happy — and label AI-made shots when you post them.
FAQ
- Is this the real GPT Image 1.5?
- It matches the 1.5 effect, but the engine is a newer model from the same family. GPT Image 1.5 itself is not onboarded on Loomi yet, so this tool is served by gpt-image-2.5-flare covering the same 1.5-style capabilities: precise following, targeted edits, progressive refinement. We switch to the real 1.5 model id and re-shoot the showcase as soon as it lands.
- Is it free? Does it cost money?
- You use it right inside Loomi, billed per generation in credits — no extra app download, no card binding. If you want the official channel instead, the free ChatGPT tier includes a few image generations per day, and Plus (about $20/month) is faster with clearer output.
- I only have one selfie. Group photos?
- Yes. One clean front-facing selfie is enough for trending outfit swaps and group photos. Keep the original well lit, unblocked and not over-beautified, change one variable at a time — outfit first, then background and people — and never ask for tiny distant faces in a group shot, they always go mushy.
- How do I prompt it to look truly shot?
- Write a photography brief, not art words. Use this template: subject + action + setting + device and focal length + light + texture + flaws. Example: candid iPhone snapshot, 28mm, late-afternoon market backlight, girl turning back laughing, visible pores, windswept hair, blurred passersby with grain. Always add candid, amateur, unposed — and never ultra detailed or perfect skin, the more perfect the faker it looks.
- How is 1.5 different from 1.0 and 2.x?
- GPT Image 1.5 is the Dec 16, 2025 release: much stronger instruction following, precise local repainting (inpainting) and multi-round edits without degradation versus 1.0. Against the newer 2.x generation it is officially the previous model — lighter and faster for everyday realistic photos and edits.
- Can I use it on my phone?
- Yes. It runs in the browser on your phone: pick a mode, type the description, optionally upload a reference photo, and generate. No Photoshop skills needed. Ask in plain words like 'a cat at the beach at sunset' — the more specific, the better looking.
- How do I spot an AI-made photo?
- No single method is 100%, but zoom in on fingers, teeth, earring symmetry and small sign text; check whether light direction, shadow length and reflections make sense; night shots that are suspiciously clean are a red flag. Ask for the original, Live photo or EXIF data, and run an AI detector as backup — never judge from one compressed forward alone.
Try it free