What is GPT Image 1.5 (I2I)?

GPT Image 1.5 (I2I) is an AI model (image editing) available on Artifio.ai through OpenAI. You can run it in our workspace from a single shared wallet and download your outputs.

How much does GPT Image 1.5 (I2I) cost on Artifio?

GPT Image 1.5 (I2I) starts at about $0.020 per generation on Artifio.ai. The exact cost is shown in the workspace before you confirm and can vary with the options you select.

How do I use GPT Image 1.5 (I2I) on Artifio?

Click Generate Now to open the workspace at /create/gpt-image-1-5-image-to-image. Fill in the form fields, review the live cost estimate, and run your generation.

What kind of outputs does GPT Image 1.5 (I2I) produce?

GPT Image 1.5 (I2I) is configured for image editing workflows in the image category. See the technical parameters below for supported inputs and options.

Can I try GPT Image 1.5 (I2I) for free?

Sign up free, top up your wallet from $5, and you can run GPT Image 1.5 (I2I) pay-as-you-go without a separate subscription for this model.

GPT Image 1.5 (I2I) — AI Image Generator

Name: GPT Image 1.5 (I2I)
Brand: Artifio
Price: 0.020 USD
Availability: InStock

$1 on us$1 on us + a free Gold trial to get you startedClaim your $1 + trial Terms apply

artifio.ai

GPT Image 1.5 (I2I) — AI Image Generator | Artifio.ai

About GPT Image 1.5 (I2I)

GPT Image 1.5 is OpenAI's December 16, 2025 image model, released as the new engine behind ChatGPT Images and available in the API as gpt-image-1.5. Precise editing was the launch's headline: OpenAI introduced post-production tools for granular edits, and TechCrunch's coverage highlighted the model's ability to change only what you ask for while holding the rest of the picture steady.

This entry covers the image-to-image mode, where you upload a picture and direct changes with plain-language instructions. Artifio runs it alongside 100+ other models in one workspace with a single pay-per-generation wallet.

Features

Accepts an input image plus a text instruction and returns an edited version, with support for repeated edit rounds on the same picture.
Preserves facial likeness, lighting, composition, and color tone across inputs, outputs, and subsequent edits.
Applies targeted changes rather than regenerating the whole frame, which OpenAI positioned as the main advance over GPT Image 1.
Returns edits at 1024x1024, 1024x1536, or 1536x1024 in low, medium, or high quality tiers.
Runs up to four times faster than GPT Image 1, so multi-round edit sessions stay responsive.

What people use it for

Photo retouching where the subject's face and the scene's lighting must survive the edit untouched.
Swapping, removing, or restyling a single element in a product shot without disturbing the rest of the frame.
Iterative creative direction: upload, adjust, review, adjust again, with each round building on the last.
Producing portrait and landscape crops of an edited image for different placements.

Reported strengths

Edit precision is the model's documented specialty; TechCrunch described consistency in likeness, lighting, composition, and color tone as the core improvement.
Iteration consistency tackles the drift problem that made earlier image editors repaint faces and backgrounds unprompted.
The fourfold speed gain over GPT Image 1 shortens the feedback loop that editing workflows depend on.
Image output tokens cost less in OpenAI's API than GPT Image 1's did.

Reported limitations

Scenes with multiple faces remain a documented weak spot for the GPT Image family, which matters more in editing where real people are usually in frame.
In-image text in Chinese, Arabic, and Hebrew is still unreliable, so edits that touch signage or labels in those scripts can fail.
Edited output is capped at 1536 pixels on the long edge, below the source resolution of most modern photos.
OpenAI's API documentation marks gpt-image-1.5 as deprecated now that GPT Image 2 handles both generation and editing.

Sources

GPT Image 1.5 Model (OpenAI)
GPT Image (Wikipedia)
OpenAI continues on its 'code red' warpath with new image generation model (TechCrunch)