Realtime text to image For now i need good prompts or model, because stable diffusion do kinda bad generations