
Oryginalny dokument
GPT Image 2 · en
Scene:
Using the referenced host identity and the same thumbnail template, create a square short-video cover for an urgent AI breakthrough story. Preserve the close face crop on the left, the empty hook block on the right, and the narrow category strip at the top. Background should remain dark navy and white, but introduce a bright electric-cyan neural diagram glow behind the host to signal AI without making the frame messy.
Subject:
The same recurring female host from the reference image appears with matching haircut, blazer, skin tone, lighting direction, and camera angle. Her expression shifts to alert concentration, with one hand slightly raised near the chin as if reacting to breaking news. On the right side, abstract non-readable headline bars sit above a simple chip icon and a glowing circuit motif. The design must look like a second episode in the same channel system, not a different brand.
Important details:
Identity consistency is mandatory: same host face, same overall crop, same channel styling, same clean geometry. The only major changes should be the cyan accent, the AI iconography, and a stronger sense of urgency in the expression. Keep the design optimized for mobile-size scanning with clear contrast and minimal clutter. All text areas must remain unreadable placeholder shapes, never actual words.
Use case:
Short Video Cover Pack series study 2/9, using the reference host to test AI-news category variation within one stable thumbnail system.
Constraints:
No readable text, no watermark, no logos, preserve exact 1:1 aspect ratio, no different host identity, no extra panel collage, no overcomplicated tech background, no soft low-contrast thumbnail.Jak używać
- Skopiuj cały oryginalny prompt wraz z parametrami.
- Dostosuj temat, tekst i kompozycję do swoich potrzeb.
- Użyj zgodnego narzędzia zewnętrznego; sprawdź warunki dostępu i wynik.
Przed rozpoczęciem
- Sprawdź, czy potrzebny jest obraz referencyjny. Zgodność nie potwierdza modelu użytego do przykładu.
- Sprawdź nazwy, liczby i napisy; wyniki zależą od wersji modelu.





