Most people still think of OpenAI’s Sora as a platform made exclusively for video generation. But recently, Sora has stepped boldly into the world of image generation — and not just with competence, but with excellence. Its image quality is striking, especially in photorealism, where it competes head-to-head with Midjourney and Flux, sometimes even surpassing them.
Sora steps into the game
Over the years, I’ve worked mostly with Midjourney, Stable Diffusion, and Flux — all in pursuit of photorealism, the aesthetic I resonate most with. When Sora was added to my ChatGPT Plus subscription, I didn’t pay much attention at first. After all, it was branded mostly as a video tool — something that doesn’t currently appeal to me. Personally, I believe video generation still has a long way to go before it reaches the finesse of image models.But then I started noticing something: some images on Sora’s feed were really good. There was a precision, realism and a depth. So I tried it. And the results? Honestly, surprising in the best possible way. Let me explain.
Depicting Plus-Size Bodies — Sora Takes the Cup
If you’re familiar with Reflexorama or any of my projects, you’ll know that the celebration of plus-size bodies is central to my work. And you’ll also know that these bodies are among the most difficult to depict in AI — not because they’re inherently difficult, but because of how society (and algorithms) view them.
Despite increasingly complex and descriptive prompts, most platforms either censored plus-size subjects or rendered them grotesquely or comically. While progress has been made — especially in Midjourney — we’re still not at the point where curvatures, softness, and humanity are portrayed with accuracy and dignity.
Sora however, is capable of rendering plus-size individuals in exquisite, believable, and compassionate detail. The folds, the postures, the warmth — all realistic. This is what body positivity should look like in AI: the return of voice, the restoration of dignity, the celebration of diversity.
Capturing Multiple Subjects in One Frame — Another Win for Sora
Another long-standing issue in AI image generation is depicting multiple distinct individuals in a single scene. Stable Diffusion 3.5 recently made some progress here, allowing different people to be rendered together, but image quality suffered — and many users have since abandoned it.
Sora, however, nails it. Not only can it place several people in the same frame, but it differentiates between them — across body types, ethnicities, and expressions. There’s no “copy-paste” look. Instead, each subject feels unique and real. That’s a huge leap forward.
Censorship and Sora
Of course, nothing’s perfect — and Sora has its own share of censorship. I expected it to be heavily filtered, but to my surprise, it’s actually quite flexible with good prompting. You won’t get artistic nudes as easily as with Stable Diffusion or Flux (especially not those enhanced with LoRAs), but Sora does understand intention. It distinguishes between beauty and vulgarity in nuanced ways.
Unfortunately, censorship still tends to be stricter when dealing with plus-size bodies — a pattern I’ve addressed across my projects. But in Sora, that wall feels like it’s beginning to crack.
Prompting Sora
I’m still learning the quirks of Sora’s prompt structure, but one thing is clear: it’s smart. With the backing of GPT’s language models, Sora understands context better than any other platform I’ve used. My advice? Prompt as if you’re talking to ChatGPT. Explain what you want clearly and conversationally — and you’ll be amazed by the results. That said, traditional descriptive prompting still works well too.
Pricing and Access
Right now, if you have a ChatGPT Plus subscription, you get unlimited image generation via Sora — two simultaneous generations, no credit systems. Sure, the wait can vary, but I don’t find it slow or limiting. The standalone Sora plan offers faster video features, but if you’re focused on still images like I am, the ChatGPT route is more than enough.
See It for Yourself
I’ve just updated Reflexorama with dozens of new images and albums made using Sora. Check out the series including Velvet Touch (I and II), Embrace, Timeless Elegance, Beauties, Seaside Rendez-Vous, The Sun of Egypt, Heavy Glories, Heat Wave and Cotton and Rust, along with the new Conceptual VII, Conceptual VIII, Conceptual IX, Vintage IV, and Mixed IV. You’ll see everything we’ve talked about: the softness, the clarity, the inclusion, the diversity — all handled with care and realism.
For me, this is a revolution. Let’s just hope OpenAI continues supporting this extraordinary feature without quietly phasing it out. Sora is a dream lens — and it deserves to stay open and accessible to all pockets.
Website: