๐จ Image Agent - Character-Aware Image Generation
Space 2/3 - Images match scene characters: ๐ค human / ๐พ animal / โจ fantasy
Device: CPU | Text Agent: โ ๏ธ Not Connected | Video Agent: โ ๏ธ Not Connected | Translation: Groq=โ + LocalFallback=โ ๏ธ keyword | ๐ญ Character-Aware: ON
10 30
๐น Final Video
โจ Character-Aware Image Generation (NEW)
How it works:
- Space 1 detects characters and classifies them as human/animal/fantasy/object/none
- Space 2 reads the
charactersarray from each scene - Adds the correct type hint to the image prompt:
| Type | Added to prompt | Negative |
|---|---|---|
| ๐ค human | "realistic human beings, photorealistic people" | "animals, creatures" |
| ๐พ animal | "realistic animals, detailed fur/feathers" | "people, humans" |
| โจ fantasy | "fantasy creatures, magical beings" | โ |
| ๐ none | "landscape focus, no characters" | "people, animals" |
- Character descriptions from Space 1 are injected at the start of the prompt
- This ensures the image model generates the correct character types for every scene
Translation: Groq=โ + LocalFallback=โ ๏ธ keyword