Users are encountering difficulties generating images with accurate spatial awareness and consistent character height differences using current text-to-image models like Flux2Klein 9B and Qwen 2.1 Turbo within the ComfyUI interface. Despite attempts to use ControlNet and detailed textual prompts specifying relative heights and positions, the models either ignore the instructions or produce exaggerated scale differences, resulting in characters that are too small or maintaining uniform sizes. The user is seeking recommendations for alternative models that can better handle these spatial nuances through natural language instructions. AI
IMPACT Current text-to-image models lack reliable spatial awareness, hindering precise control over character dimensions and relationships in generated images.
RANK_REASON User is asking for alternative tools/models for a specific image generation problem.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
