OpenAI has quietly begun rolling out GPT-Image-2 to paid ChatGPT subscribers, with no announcement, no blog post, just a significantly more capable image model showing up in Plus and Pro accounts starting April 19, 2026.

What Happened

GPT-Image-2 entered silent deployment to ChatGPT Plus and Pro users on April 19, according to reports from multiple paid subscribers. OpenAI has not issued an official announcement or confirmed the rollout. The model is described as a "preliminary release for model adjustment and fine-tuning purposes", OpenAI's standard approach to staged testing before a formal launch.

This follows months of buildup. The model first appeared on the LM Arena evaluation platform in early April under the codenames maskingtape-alpha, gaffertape-alpha, and packingtape-alpha before being pulled within hours. That earlier leak confirmed near-perfect text rendering, a longstanding weakness of AI image generators. The April 19 deployment marks the first time real users are getting consistent access.

Why It Matters

The headline capability is storyboard generation: GPT-Image-2 can produce 3x3 image grids showing a single character across different actions, expressions, and camera angles, with consistent facial features throughout. This was previously a signature strength of Google's Gemini image model, which dominated professional storyboard workflows. A capable, integrated storyboard tool inside ChatGPT removes a key workflow step for video creators, concept artists, and game designers who currently bounce between multiple tools.

The timing also matters. DALL-E 3 shuts down on May 12, 2026. GPT-Image-2 is the intended replacement, and a pre-launch deployment to paid users gives OpenAI time to catch bugs before the full migration window opens.

Key Details

  • Available to ChatGPT Plus and Pro subscribers as of April 19
  • No official launch announcement as of April 20
  • Text rendering accuracy reported at over 99% on standard typography benchmarks
  • Storyboard generation: character-consistent 3x3 grids in 16:9 format
  • Described as matching or exceeding Google Nano Banana Pro on storyboard tasks
  • GPT-Image-1.5 remains the current API version; GPT-Image-2 API access is expected to follow
  • DALL-E 3 retirement deadline: May 12, 2026

What to Do Next

If you have a ChatGPT Plus or Pro subscription, you may already have access: test it by prompting for a character reference sheet or storyboard sequence and see whether the output shows improved consistency. Results will vary during this staged rollout, so not every Plus account has received the upgrade yet.

For teams currently building image workflows around DALL-E 3 or GPT-Image-1.5, the API migration path is still unclear. Watch for an official OpenAI announcement before switching production pipelines. The May 12 DALL-E 3 deadline gives roughly three weeks to plan.