Studio workspace
Generate images with GPT Image 2.5 on one light table. Route Flare when you need speed, Sunburst when a frame has to hold, then export 4K or alpha.
Overview
GPT Image 2.5 is OpenAI’s image model, released in September 2026. In ChatGPT it ships as ChatGPT Images 2.5. It follows prompts and reference photos more tightly than Images 2.0, keeps subjects stable across edits, and renders light and materials closer to a camera.
The API exposes two IDs: gpt-image-2.5-flare and gpt-image-2.5-sunburst. Emulsion is a studio workspace, so you pick Flare or Sunburst per job instead of locking the whole account to one model. Latency is down by up to 50% versus Images 2.0. That matters for campaign stills, product packs, fashion looks, and interiors that need a real revision loop — not a one-shot chat sketch.
Use GPT Image 2.5 when you need production output: 4K frames, opaque or transparent backgrounds, and a sheet with history. See the example frames, compare Flare and Sunburst, then get GPT Image 2.5 access.
Gallery
These frames are the jobs studios actually run — portrait, product, fashion, and space. Keep them on a sheet, then export.
Workflow
A job has three steps on the light table: prompt or reference, choose a bath, then revise and export.
01
Start from words, a sketch, or a still. The model keeps that frame on the sheet.
02
Flare for volume and comps. Sunburst when an edit still has to look like the same picture.
03
Change a prop, hold the face, then download PNG, JPEG, or WebP from the job.
Models
OpenAI split GPT Image 2.5 into two baths. One is for speed. One is for control. Pick per job on the same sheet.
gpt-image-2.5-flare
Speed first. GPT Image 2.5 Flare stays close to GPT Image 2 quality, with better edits and subject hold, and latency down by up to 50%. Use it for social, prototypes, and volume.
gpt-image-2.5-sunburst
Precision first. Sunburst takes longer in exchange for tighter light, materials, and instruction following. Use it when a revision still has to look like the same picture.
Craft
Speed
Versus Images 2.0, GPT Image 2.5 cuts latency by up to 50%. Sketching a frame should not feel like waiting on a tank.
Edit
The model is stronger on reference fidelity and multi-turn edits. Swap a background or a prop without the subject drifting.
Light
Natural light, materials, and texture are the point of this release. The sheet is graded to that standard — no plastic render look.
Scale
The API covers 1:1, 3:2, portrait, 2K, and 4K. Transparent backgrounds export as PNG or WebP for commerce and layout.
Pricing
Reserve GPT Image 2.5 access now. We bill in USD only after you get a login. Cancel from the same email thread.
Proof
Free
to start
Learn GPT Image 2.5 on the sheet. Flare only, watermarked previews.
Studio
$29
/ month
The GPT Image 2.5 bench for campaigns and product stills.
Bench
$79
/ month
Shared queue for a small team on one light table.
FAQ
GPT Image 2.5 is OpenAI’s image model from September 2026. In ChatGPT it is labeled ChatGPT Images 2.5. It improves lighting, texture, subject hold, and multi-turn edits versus Images 2.0, with latency down by up to 50%. The API variants are Flare and Sunburst.
Flare (gpt-image-2.5-flare) is built for speed — social, prototypes, volume. Sunburst (gpt-image-2.5-sunburst) is built for control — campaigns, product shots, longer generation. Emulsion lets you switch GPT Image 2.5 models per job.
Join the waitlist. We open seats in waves and email a Studio login in queue order. There is no charge until your seat is live. Reply to opt out anytime.
GPT Image 2.5 supports common frames from 1024² through 2K and 4K, plus opaque or transparent backgrounds. Transparent output defaults to PNG or WebP.
ChatGPT is for a single sketch. Emulsion is a production workspace: prompt history, per-frame management, Flare/Sunburst routing, and export. It uses the official Images API — not a ChatGPT wrapper.
No. Emulsion is independent and calls GPT Image 2.5 through OpenAI’s Images API. GPT Image, ChatGPT, and OpenAI are trademarks of OpenAI.
Early access
Join the GPT Image 2.5 waitlist. Seats open in waves. You get a Studio login, then we bill the plan you reserved. No charge to join the list.
