AIREITERGPT Image 1.5: OpenAI's fast model for precise image edits
OpenAI released GPT Image 1.5 (API model gpt-image-1.5) in December 2025 together with the new ChatGPT Images. GPT-Image 2 and 2.5 have shipped since, and 1.5 is still available in the API for text-to-image and image-to-image work.
Up to 4x faster generation
OpenAI's launch post says GPT Image 1.5 generates images up to four times faster. That leaves room for more candidates per session while you are still exploring a direction.
Edits that change only what you ask
In image-to-image edits the model follows the request down to small details and keeps lighting, composition and people's appearance consistent from one edit to the next. Swap a jacket, remove an object or change the background without redrawing the whole frame.
Better dense text
OpenAI lists improved dense text rendering among the launch changes, which helps posters, menus, packaging and simple infographics. Put exact copy in quotes and proofread the result before it ships.
GPT Image 1.5, GPT-Image 2 or GPT-Image 2.5?
All three are OpenAI image models on AIReiter. Pick by output size, aspect ratio range and how much an edit must hold.
GPT Image 1.5 for fast, high-volume work
Drafts, social posts, product mockups and batch jobs where 1:1, 3:2 or 2:3 output is enough. Keep quality on medium for rounds you will throw away.
GPT-Image 2 for size and ratio range
GPT-Image 2 on this site outputs 1K to 4K with auto plus 13 fixed aspect ratios, including wide 16:9 and 21:9 frames. Use it when a brief needs a banner, a phone wallpaper or a print-size file.
GPT-Image 2.5 when edits must not drift
In the newer 2.5 family, Sunburst is built for edits that must leave everything else locked, and Flare is the fast default. Run the same prompt and references on 1.5 first and switch only if the result improves.
Text-to-image and image-to-image use cases
Virtual try-on and outfit swaps
Upload the person and the clothing images together, then describe which garment goes where. Name what must stay the same: face, hairstyle, pose and background.
Product shots and packaging variants
Start from one product photo and change a single detail per request, such as the label color, the flavor banner or the scene behind the pack. Medium quality is usually enough for review rounds.
Posters and social graphics with copy
Write the headline in quotes, describe the layout and pick 2:3 for a poster or 1:1 for a feed post. Switch to high quality for the final version.
Multi-image composites
Combine up to 16 reference images for a mood board, a character in a new setting or several products in one scene. Say which image supplies the subject and which supplies the style.
Medium or high quality
Quality is the main trade-off: medium for speed, high for detail. Credits for each tier are listed in the pricing table on this page.
Medium: fast and balanced
The default. Use it for exploration, A/B tests and anything you will generate many times before choosing one.
High: slower, more detail
Takes longer to render. Use it for the final version of a composition you have already settled on with medium.
Aspect ratio and references
Output is 1:1, 3:2 or 2:3. Reference images are optional: add them for image-to-image edits and composites (JPEG, PNG or WebP, up to 10 MB each, up to 16 per request), or leave them out for text-to-image.
How to use GPT Image 1.5
From prompt to finished image in the playground above.
Write the prompt
Describe the image, or for an edit, the one change you want and what must stay the same.
Add references for image-to-image
Upload up to 16 images. Without references the request runs as text-to-image.
Pick ratio and quality
Choose 1:1, 3:2 or 2:3, then medium for drafts or high for the final.
Generate and iterate
Keep the result you like and describe the next change. Failed tasks are not charged.
Other image models to compare
GPT-Image 2
OpenAI image model with 1K to 4K output, auto plus 13 fixed aspect ratios.
GPT-Image 2.5
OpenAI's newer family: Flare for speed, Sunburst for edits that must hold.
Nano Banana Pro
Google's image line for editing and multi-image composition up to 4K.
Seedream 5.0 Pro
ByteDance stills with large reference sets.
GPT Image 1.5 FAQ
Supported options, billing, availability and how it compares with GPT Image 1 and newer GPT image models.
/ 01What is GPT Image 1.5?
It is OpenAI's image generation and editing model released in December 2025 with the new ChatGPT Images, the successor to GPT Image 1. It creates images from text and edits or combines reference images.
/ 02Is GPT Image 1.5 still available?
Yes. OpenAI has since released GPT-Image 2 and GPT-Image 2.5, and gpt-image-1.5 remains available in the API. On AIReiter it runs in this playground and through the API.
/ 03How is GPT Image 1.5 different from GPT Image 1?
According to OpenAI, 1.5 generates up to four times faster, follows edit instructions more closely while keeping lighting, composition and people's appearance consistent, and renders dense text better.
/ 04How is GPT Image 1.5 billed?
Per image, by quality tier. Credits for medium and high are listed in the pricing table on this page. Text-to-image and image-to-image are billed the same way, and failed tasks are not charged.
/ 05Which aspect ratios and inputs are supported?
Output is 1:1, 3:2 or 2:3. You can attach up to 16 reference images in JPEG, PNG or WebP, up to 10 MB each. Each request returns one image.
/ 06Does GPT Image 1.5 support transparent backgrounds?
Not on AIReiter. This playground and the API expose prompt, reference images, aspect ratio and quality only, so outputs have an opaque background. For a cut-out, generate on a plain background and remove it in a separate step.
/ 07Is GPT Image 1.5 better than GPT-Image 2?
Not across the board. GPT Image 1.5 is fast with two simple quality tiers, while GPT-Image 2 on this site offers 1K to 4K output and more aspect ratios. Run the same prompt on both and keep the one that wins for your brief.
/ 08Can I use GPT Image 1.5 through the API?
Yes. OpenAI's model ID is gpt-image-1.5; on AIReiter, send model gpt_image_1_5 to POST /api/openapi/submit and poll the task for the result. Parameters are prompt, image_url (optional, up to 16 for image-to-image), aspect_ratio and quality.