Task: Photorealistic cinematic portrait with correct hands and faces.
For portrait realism at the top of the market, Midjourney v6.1 and FLUX 1.1 Pro are nearly tied — Midjourney edges on skin texture and color grade, FLUX edges on hands and literal prompt following. Gemini Imagen 3 is a legitimate free alternative for 90% of use cases. Stable Diffusion is only worth it if you need unlimited local gen.
Cinematic portrait of a 32-year-old south-asian man, salt-and-pepper beard, wearing a dark grey wool coat, looking slightly off-camera. Shot on Sony A7III with a 50mm f/1.4 prime lens, moody natural lighting from a window at golden hour, shallow depth of field, filmic color grading, 35mm film grain, hyperrealistic skin texture with visible pores, photographed in Mumbai Film City, aspect ratio 4:5.
| Rank | Model | Total /50 | Cost / run |
|---|---|---|---|
| 🥇 | Midjourney v6.1 | 48 | 4.0¢ |
| 🥈 | FLUX 1.1 Pro | 47 | 5.0¢ |
| 🥉 | DALL-E 3 | 39 | 4.0¢ |
| #4 | Stable Diffusion 3.5 | 39 | 1.0‰ cent |
| #5 | Gemini Imagen 3 | 37 | Free |
Best overall quality. Perfect skin texture, correct 10-finger hands, color grade matches the "filmic" ask exactly.
Winner on hands and prompt adherence — the dark grey wool coat was spot-on. Face slightly softer than Midjourney's.
Best at following EXACT prompt (included the Mumbai location implicitly). Skin texture plasticky. Hands OK.
Needed 2 regens for correct hands. Good filmic grade but less "wet-film" realism than MJ or FLUX.
Free is the key here. Good face, OK hands, slightly lost the "golden hour" lighting ask.