This paper presents a training-free baseline framework for text-to-face generation using pretrained Stable Diffusion and CLIP models. No additional model training or fine-tuning is performed. The proposed approach focuses on facial-specific prompt-engineering strategies and CLIP-based semantic evaluation to study text–image alignment in facial synthesis. Experimental evaluation on a limited set of facial descriptions demonstrates consistent semantic similarity with a CLIP score of 0.651, with an average generation time of under 15 seconds per image. Rather than competing with state-of-the-art methods, this work is intended as a reproducible baseline to support further research in semantic-driven facial image synthesis.