How to generate AI images using ChatGPT (www.engadget.com)

OpenAI’s ChatGPT now generates and edits images directly inside the chat interface for free users, restoring broader access after a brief restriction. Users can create images from text prompts or upload existing photos to guide generations and edits — including inpainting/masking via a paintbrush tool, iterative “try again” options, and mobile/desktop upload flows. Outputs are non-deterministic (same prompt can yield different results), can take up to a couple of minutes depending on server load, and the service enforces copyright guardrails (it won’t reproduce exact real-world photos). Users can opt out of having uploaded images used to train models through account Data controls. This matters for the AI/ML community because it further democratizes multimodal generation and makes in-chat image editing a mainstream workflow, lowering the barrier for rapid prototyping of visuals and UX-driven image manipulation. Key technical implications: integrated inpainting and prompt-driven editing accelerate creative iteration but expose typical generative-model issues (hallucinated composition, inconsistent details); server-side rate limits and latency affect throughput; and user opt-outs and copyright restrictions impact the availability of training data. OpenAI also tiers performance—Plus ($20/mo) for faster/expanded creation and Pro ($200/mo) for unlimited generation—so cost and reproducibility remain important considerations for power users and researchers.