Eight Neural Networks Drew Bangs, Three AI Stylists Advised Against Them: How to Try on Hairstyles and Looks via Photo

The author conducted an experiment on choosing hairstyles using neural networks, dividing the process into two stages: consultation and visualization. First, language models (Claude, GPT, Gemini) analyze a photo and provide stylistic recommendations, considering face shape and accessories. Then, specialized image editing models visualize the suggested changes. Testing eight models via the BotHub aggregator revealed that gpt-image-2.5 offers the best price-to-quality ratio while preserving original facial features. The author notes that AI stylists can be inconsistent in determining face shapes, a fact supported by studies on human subjectivity in appearance assessment. Using a combination of an LLM for advice and an editing model for visualization allows for effective look testing, though it requires a critical approach to AI outputs.