GPT Image 1.5: ChatGPT's New Creative Studio in Your Pocket

OpenAI's GPT Image 1.5 now powers both the refreshed ChatGPT Images experience and the API. The model emphasizes 4× faster generations, tighter control over edits, and richer creative tooling, effectively turning ChatGPT into a portable studio for image experimentation.

Results That Actually Match Intent

GPT Image 1.5 is trained to preserve what matters—lighting, composition, likeness, and other delicate context—while changing only the requested elements. When users upload a photo and describe a tweak, the model keeps backgrounds and faces consistent even across multi-step edit sessions, making hairstyle try-ons, wardrobe swaps, and stylistic filters more reliable.

Precise, Multi-Step Editing

OpenAI showcases edit chains where candid photos merge into retro 2000s film shots, chaotic party guests appear in the background, one subject becomes hand-drawn anime, the dog turns into a plushie, and the scene eventually morphs into a livestream set. Each instruction builds on the previous one without erasing earlier intent, illustrating how the model adds, subtracts, blends, and transposes elements while guarding the core composition.

Creative Transformations With or Without Prompts

Transformations that remap scenes, text, and layout now retain important details. Examples span "codex"-themed movie posters with costume swaps, glam-doll fashion ads, and ornamental layouts. The new ChatGPT Images interface supplements prompt-based control with preset styles, so experimentation can happen even without typing a request.

Instruction Following That Holds Structure

Complex compositions—such as 6×6 grids populated with specific objects per cell—render accurately. That fidelity extends to original generations where spatial relationships matter, helping storyboards, product lineups, and other structured scenes maintain the intended arrangement rather than drifting.

Text Rendering That Finally Looks Right

Dense markdown layouts, calorie infographics, and other typography-heavy visuals appear cleaner. GPT Image 1.5 packs smaller fonts into tall compositions (like the GPT‑Image‑1.5 newspaper mockup) without the smearing or missing glyphs that often plagued earlier models, which benefits UI mocks, editorial content, and data-rich graphics.

A Faster, Dedicated Creation Space

The new Images surface in ChatGPT's sidebar (web and mobile) introduces preset filters, trending prompts, and a one-time likeness upload, so regular collaborators can reuse their appearance without digging through camera rolls. Rendering speed increases by up to four times, and additional prompts can be queued while earlier generations finish, which keeps creative exploration moving.

Quality Gains and Known Limits

Side-by-side comparisons highlight sharper outputs in challenging contexts. Crowded 1970s London street scenes keep dozens of small faces distinct, glare-heavy photos maintain realism, and deep-sea posters capture more scientific detail—though OpenAI notes that accuracy is still a work in progress. Scenarios with numerous faces, certain styles, or multilingual text continue to expose limitations, but the trajectory shows clear improvement over GPT Image 1.0.

API and Enterprise Readiness

GPT Image 1.5 ships in the API under the gpt-image-1.5 name and carries the same preservation and editing strengths as the ChatGPT experience while cutting image input and output prices by roughly 20% compared to GPT Image 1. The company cites adoption from Wix, Canva, Higgsfield, Figma Weave, Envato, and other creative and commerce platforms that need consistent logos, lighting, and fine-grained detail across product shots, marketing assets, and catalog variations.

Availability and Practical Workflows

The upgraded model rolls out across ChatGPT surfaces automatically, while the earlier image experience remains accessible as a custom GPT for anyone who prefers it. Typical workflows look like this:

1. Teams open ChatGPT and use the Images sidebar to browse preset filters or upload a likeness for reuse.
2. Editors chain multiple instructions—mixing style changes with additive and subtractive edits—to watch the model respect prior context without rewriting the scene.
3. Developers hit the API or Playground to generate branded assets at scale, benefiting from lower pricing to iterate through more product or marketing variations.

Together, these upgrades make GPT Image 1.5 feel like a dependable creative partner: faster renders reduce wait times, instruction fidelity keeps edits on brief, text rendering finally reads cleanly, and both consumers and enterprises gain a clearer path from prompt to polished visual.