December 16, 2025
Banana beef and pixel drama
GPT Image 1.5
GPT Image 1.5: Faster pics, sharper edits — but artists aren’t impressed
TLDR: OpenAI’s GPT Image 1.5 promises faster, cheaper, more precise image edits with a new in-app Images hub. Commenters aren’t sold: Midjourney fans call it too photoreal, tinkerers compare it unfavorably to the “Nano Banana” benchmark, and a verification glitch added to the chaos—style over speed is the debate.
OpenAI just dropped GPT Image 1.5, promising speed (up to 4x faster), cheaper image runs (20% off), and cleaner edits that keep lighting, composition, and faces intact. There’s even a new Images hub inside ChatGPT with preset styles to spark ideas. The official pitch: a “creative studio in your pocket.” The community pitch: hold that confetti. On the HN thread sharing the announcement, the vibe is equal parts “nice” and “nah.”
The strongest heat comes from the art crowd. One user declares they’re sticking with Midjourney, saying the big players chase photoreal vibes while style takes a back seat. Meanwhile, the open-source testers rolled in with the running joke/benchmark du jour, “Nano Banana.” One well-known tinkerer said GPT Image 1.5 is better than Nano Banana… but “MUCH worse” than their top bar, leaving readers cackling at the banana-meter drama. Another critic poked at OpenAI’s own curated samples, saying they still struggle with consistent edits.
Adding to the chaos, a user got trapped in a verification limbo—a green checkmark with nowhere to go—before it “mystically” fixed itself. So yes, the new model can preserve logos and render tiny text, but the thread’s mood? Show us the style, not just the stock-photo polish.
Key Points
- •OpenAI released GPT Image 1.5 and a new ChatGPT Images feature for faster, more precise image generation and editing.
- •GPT Image 1.5 improves adherence to user instructions, preserves details, and enhances text rendering and multi-face quality.
- •A dedicated Images home in ChatGPT offers preset filters and prompts, enabling exploration without written prompts.
- •Image inputs and outputs are priced 20% lower than GPT Image 1, and the model generates images up to 4x faster.
- •The release targets practical and creative use cases, including marketing, logo creation, and e-commerce product catalog generation; prior ChatGPT Images remains as a custom GPT.