DALL-E 3

The 200 user comments (primarily from HackerNews) reveal a deeply split reality. On the positive side, non-artists report transformative accessibility: 'I can write a prompt to get me a good-enough image in 90% of cases whereas before that was totally unreachable without an artist.' The raw aesthetic quality draws awe — 'the outputs on the page are just awesome. Leagues ahead.' However, power users flag critical limitations. Multiple commenters note that OpenAI silently rewrites prompts to inject diversity directives: 'openais internal prompt for dalle modifies all prompts to add diversity and remove requests to make groups of people a single descent.' This hidden prompt layer means users don't get full transparency on what the model actually receives. Rate limits are a major friction point: 'you can get hit by 5 img/min limits, or 100 RPD (requests per day).' The image-to-text feedback loop is described as broken: 'the image-to-text model doesn't have any idea what the pictures actually contain... it's just matching patterns.' Experienced users consistently point to Stable Diffusion + ControlNet + LoRA as superior for precise structural control and artist-style replication: 'with controlnet you can very precisely specify how your image should be structured. This is closer to how traditional artists work.' Prompt engineering fatigue is real: 'I find the generative art bits really frustrating, and typing in long prompts to be also kind of not worth the effort, I can usually just make what I want in less time than it takes to refine my prompts down.' Broader sentiment is bittersweet — admiration for the technology tempered by existential anxiety about the future of commercial art.

No aggregate ratings were found for this product during the last harvest.