Нейронки


Гео и язык канала: Россия, Русский
Категория: Технологии



Гео и язык канала
Россия, Русский
Категория
Технологии
Статистика
Фильтр публикаций


Репост из: лепроиздим
SYSTEM ROLE
You are an experimental studio photographer and CGI art director specializing in extreme wide-angle imagery, exaggerated perspective, and sculptural editorial compositions.
MAIN TASK
Transform the subject from the attached photograph into a dramatic ultra-wide-angle studio image. Generate ONE image featuring ONE subject. Use the attached photograph to preserve the subject’s identity and appearance, while rebuilding the camera position, perspective, framing, and lighting.
SUBJECT PRESERVATION — CRITICAL
Keep the subject recognizable. Preserve its defining shapes, colors, materials, construction, and existing details. For a person or character, preserve their identity and outfit.
Preserve the subject’s actual structure while allowing strong optical exaggeration of its apparent proportions. The foreground must look oversized because it is extremely close to the lens.
CAMERA AND PERSPECTIVE — CRITICAL
Create the visual effect of a 10–14 mm full-frame ultra-wide-angle lens placed extremely close to the subject.
Push the nearest distinctive part toward the camera so that it becomes disproportionately large and visually dominant. Let the rest of the subject rapidly diminish into the distance. Produce a striking near-to-far scale difference, strong foreshortening, and an immersive sense of depth.
The camera should feel almost inside the subject’s immediate space. Merely rotating the subject into a three-quarter view is not enough.
Allow bold wide-angle stretching and rounded foreground expansion. Keep the image full-frame, without a circular fisheye border.
VIEWPOINT AND ORIENTATION
Choose an unconventional oblique viewpoint that emphasizes the subject’s depth. Position the lens near the subject’s lower or forward extremity and look diagonally across or upward along its structure. Show upper or lower surfaces as appropriate to its shape.
Tilt and rotate the subject in three-dimensional space. The essential visual relationship is a massive near element projecting toward the viewer and a much smaller far element receding behind it.
COMPOSITION AND FRAMING — CRITICAL
Create a single bold, frame-filling composition. The subject should occupy approximately 80–90% of the image, with its silhouette clearly readable.
Place the enlarged nearest part in the lower foreground, close to the bottom edge. Let the remaining structure travel diagonally toward the upper part of the frame. Introduce a dynamic off-axis tilt.
Keep narrow white margins around the silhouette. Fit the complete subject into the frame without weakening the extreme perspective.
The subject must feel as though it is pushing out of the image toward the viewer.
BACKGROUND — ABSOLUTE REQUIREMENT
Use a completely solid, pure white background: HEX #FFFFFF, RGB 255, 255, 255.
Every background area outside the subject’s silhouette must remain the same uniform pure white, including spaces between its parts and all four corners of the image.
No off-white, ivory, cream, light gray, color tint, gradient, vignette, texture, noise, backdrop shading, horizon, or visible floor. No cast shadow, contact shadow, reflection, glow, or ambient haze on the background.
Treat the background as a flat #FFFFFF field independent of the subject’s lighting. Keep natural edge antialiasing confined to the silhouette.
SCENE
Present the subject as suspended in space, without a visible support, pedestal, or resting surface.
One subject only. No collage, grid, repeated views, inset photographs, or surrounding objects.
LIGHTING
Use a large, soft studio light above and behind the subject, with restrained frontal fill. Create luminous upper edges and broad highlights, balanced by deep, soft shading on the near-facing and underside surfaces.
Keep all lighting effects, tonal transitions, and shadows confined to the subject itself. Preserve rich midtones and sculptural volume while the background remains perfectly uniform #FFFFFF.
STYLE AND MOOD
Playful, surreal, tactile, and visually bold. Combine photographic material realism with the exaggerated spatial impact of an experimental CGI editorial image.
The subject should feel unusually close, substantial, and dimensional, with a slightly surreal sense of scale.
TEXTURE AND FINISH
Render believable surface textures and material-specific reflections. Preserve fine detail without excessive sharpening. Use gently softened highlights within the subject, without any glow spilling onto the white background.
Keep the subject sufficiently clear throughout for its shape and construction to remain readable. Create depth through perspective rather than heavy blur.
CRITICAL PRIORITIES
Completely solid pure white background: #FFFFFF.
Extreme camera proximity.
Oversized foreground element.
Rapid reduction in apparent scale toward the background.
Strong diagonal depth and foreshortening.
A floating, isolated subject.
One final image of one subject.
If the result looks like a conventional product photograph, move the virtual camera much closer and strengthen the near-to-far size difference.
DO NOT INCLUDE
A distant camera, telephoto compression, orthographic or isometric projection, a flat side view, a standard overhead product shot, timid perspective, uniform apparent scale, any background color other than #FFFFFF, background gradients, background shadows, a visible studio floor, extra parts, melted geometry, added typography, captions, watermarks, frames, or interface elements.


Репост из: лепроиз
💥Ловите еще примеры генераций с этим промптом


Репост из: лепроиз
😍Делаем пушечный угол в Nano Banano Pro

Промпт в коммах✏️


Репост из: Psy Eyes
WorldLabs: выпустили генератор миров Atlas с детальным контролем камеры и 2К на выходе.

Модель натренирована с нуля на тексте, картинках, видео, и 3D. Это мультимодальный авторегрессивный диффузионный трансформер, который берёт все входящие данные в один контекст и использует его для создания консистентных продолжительных сцен на основе сплатов.

Фичи:

* Контроль камеры: Atlas генерирует картинки и видео на основе одного и более исходных изображений с пиксельно точным управлением камерой, создавая видео продолжительностью до 1 минуты в разрешении 1440p. Можно задавать разные траектории для пролёта сквозь одну и ту же стабильную сцену.

* Пространственная реконструкция: модель реконструирует сцены и генерирует новые углы обзора на основе исходных картинок/видео без необходимости в специализированном оборудовании. Можно сделать шот на телефон(ы) и дальше вытаскивать из него дубли с разных ракурсов. Полученные результаты можно скачать как видео или сплат для дальнейшей работы со сценой в 3D софте.

* 4D симуляция: с помощью Atlas можно не просто продлять вкинутое видео, а двигаться сквозь него во времени и менять положение камеры как хочешь. При этом композиция сцены и общая детальность остаются близкими к оригиналу. Можно вкидывать и не видео, а одну картинку или больше.

* Генерация изображений: также модель по тексту генерирует изображения в разных стилях и 360-градусные панорамы разной сложности.

Ближайшее 4D-шное, что можно заценить эту тему это видосы у 4dv.ai, но там видны сплатные артефакты, а у Atlas сцены чистые.

Пока Atlas за вейтлистом. Подаём заявку - кто-нибудь из нас да пробьётся.

Анонс
Сайт
Вейтлист

Показано 4 последних публикаций.