Actuellement, j'utilise quatre algorithmes d'Intelligence Artificielle pour générer des images. J'ai proposé les mêmes prompts à chacun d'eux, sachant que "Ideogram.ai" et "Bing Dall-E" n'utilisent pas de prompts négatifs.
Prompt positif : "(((handmade ink drawing))), a beautiful Balinese rachel wood hugging emotionally her young daughter, wearing delicate Balinese batik, delicate and intricate Balinese pattern as background, high contrast, watercolor, centered, dynamic pose, extremely detailed, sharp focus, hand-drawn, 2D, two-dimensional, concept art, (sepia:1.1) (comic style), vector, (flat colors:1.5), elegant , National Geographic photography, photorealistic, hyper realistic , shot on Hasselblad X1D 100MM, natural lighting, emotional face".
Prompt négatif : "(shading:1.6) (3D:1.6) (volumetric:1.6) (colors) (skin color:1.6) (skin:1.6) (realistic) (red:1.6) (pink:1.6) (yellow:1.6) (green:1.6) (grey:1.6) (black:1.6) (blue:1.6) (purple:1.6) low saturation, nudity, boobs, nipples, doll, soft colors, poorly detailed face, flying, frame, framed, elongated body, six fingers, wrong hands, four finges, ugly, bad face, jpeg artifact, poorly drawn face, two scenes, double scene, incorrect hands, malformed hands, blurry face, extra-saturated, low details, low resolution, incoherent, tiling, poorly drawn hands, body out of frame, head out of frame, malformed body, pixelated, (poor quality:1.4), clipped parts, clipped images, unclear, cut off, draft, disjointed, poorly detailed, low details, incomplete, mutations, extra limbs, disfifured, texts, watermark, letters, artist name lowres, username, (normal quality1.4), double results, double head, polar lowres, gross proportions, weird perspective, wrong perspective, wrong sizes, (text:1.1)" .
Aucun commentaire:
Enregistrer un commentaire