The AI Image Generation Honest Review

I evaluated four AI image generation tools: Midjourney v6, DALL-E 3 (via ChatGPT Plus (See Bottley's full review →)), Stable Diffusion (self-hosted), and Adobe Firefly. My evaluation criteria: photorealistic quality, consistency across generations, prompt adherence, and commercial use licensing.

Photorealistic Quality

Midjourney v6 produces the highest photorealistic quality in the current evaluation. The gap versus DALL-E 3 is meaningful — Midjourney's lighting, texture, and face rendering are noticeably better for commercial photography use cases. Adobe Firefly's photorealistic quality is improving rapidly and is approaching Midjourney's level for product photography specifically.

Consistency

Consistency — the ability to generate similar images across multiple prompts — matters for brand and product use cases. Midjourney's consistency is good but requires explicit style parameters. Adobe Firefly's Project Concept model offers better consistency for users who need stable brand visual styles. DALL-E 3's consistency is the weakest of the three.

Commercial Licensing

Adobe Firefly uses only Adobe-licensed and public domain training data. All generated images are commercially usable without intellectual property concerns. Midjourney and DALL-E 3 have terms that permit commercial use but with less clarity on training data provenance. For brands that require clean IP chains, Firefly is the safest choice regardless of quality tradeoffs.

My Recommendation

Midjourney for highest quality where IP provenance is not a primary concern. Adobe Firefly for commercial use cases requiring clean licensing and style consistency. DALL-E 3 for users already paying for ChatGPT Plus who don't need the highest quality tier.