GPT Image 1.5: FAQ
GPT Image 1.5 is OpenAI's image model on Luma, best known for generating images on a true transparent background for cutouts, logos, stickers, and product shots.
1. What is GPT Image 1.5?
GPT Image 1.5 creates images from a text prompt and edits existing ones. Its standout feature is native transparency: it can output a real RGBA cutout with no background, which most image models cannot do. It preserves the context of input images, including style and composition, and is a good fit when you need clean compositing assets.
2. Which aspect ratios does it support?
Three: 1:1 (square), 3:2 (landscape), and 2:3 (portrait). For best results, match your input or intended output to one of these, since other shapes get cropped or resized.
3. Can I generate transparent backgrounds?
Yes, this is its strength. Set the background to transparent and use PNG (or WebP) output to get an RGBA cutout ready for compositing. You can also choose an opaque solid background, or let the model decide automatically.
4. Can I use reference images?
Yes. You can add up to 4 reference images to guide style, character, composition, or to edit. When you use more than one, label each image's role in the prompt (for example, character reference versus style reference) so the model knows what to take from where.
5. What quality and format options are there?
Quality has three levels: low (fast), medium (balanced), and high (best). Output formats are PNG (lossless, supports transparency), JPEG (smaller files), and WebP (modern compression). For transparent cutouts, stay on PNG or WebP.
6. Does it do both create and edit?
Yes. There is a create path for generating a new image from a prompt (with optional references), and a modify path for editing an existing source image with up to 4 additional references. Both support the same transparency, quality, and format controls.
At a glance
Axis | GPT Image 1.5 |
|---|---|
Modes | Create and edit |
Aspect ratios | 1:1, 3:2, 2:3 |
Transparency | Native RGBA cutouts (PNG or WebP) |
References | Up to 4 images |
Quality | Low, medium, high |
Output formats | PNG, JPEG, WebP |
GPT Image 1.5 vs GPT Image 2
Reach for GPT Image 1.5 when you need a native transparent background. GPT Image 2 does not generate transparency natively, but it takes up to 16 reference images, offers 10 aspect ratios from ultrawide to ultra-tall at 1K, 2K, or 4K, and does exact non-celebrity face copy-paste. In short: GPT Image 1.5 for clean cutouts, GPT Image 2 for wide framing, high resolution, and many references.