compressed ChatGPT Image GPT-4o prompts

5 months ago

A wide image taken with a phone of a glass whiteboard, in a room overlooking the Bay Bridge. The field of view shows a woman writing, sporting a tshirt wiith a large OpenAI logo. The handwriting looks natural and a bit messy, and we see the photographer's reflection. The text reads: (left) "Transfer between Modalities: Suppose we directly model p(text, pixels, sound) [equation] with one big autoregressive transformer. Pros: * image generation augmented with vast world knowledge * next-level text rendering * native in-context learning * unified post-training stack Cons: * varying bit-rate across modalities * compute not adaptive" (Right) "Fixes: * model compressed representations * compose autoregressive prior with a powerful decoder" On the bottom right of the board, she draws a diagram: "tokens -> [transformer] -> [diffusion] -> pixels"

1 month ago

Prompt: Realistic candid photo of Brisbane city, Australia, 410 Queen Street CBD, 10 am in summer, shot on an iPhone 10 from the footpath. Bigfoot is escaping out of the back of a parked white truck in the middle of the intersection, looking startled, wild, angry, and confused. Pedestrians are gathered at the sides of the street, shocked and panicked, some mid-movement with raised arms or stepping back. The scene has harsh midday sunlight with strong shadows, slight overexposure, imperfect framing with a slight tilt, handheld camera shake, washed-out colours, subtle photo noise, random pedestrians, cars, street signs, and street clutter. Bigfoot must look like a raw, accidental snapshot subject: slightly out of focus, with motion blur on the limbs, washed-out highlights on patches of fur from harsh sun, uneven lighting, rough edges, imperfect proportions, and candid movement. The creature blends into the noisy environment with the same compression artefacts, imperfect detail, and rough snapshot quality. The image should look like an unedited, shaky iPhone capture — not staged, cinematic, or polished. Negative Prompt (force realism): no CGI, no 3D render, no cartoon, no illustration, no painting, no fantasy art, no staged composition, no cinematic lighting, no studio lighting, no HDR, no smooth skin or polished fur, no glowing effects, no futuristic objects, no symmetry, no fake textures, no perfect sharpness, no plastic look.