Penlify Explore Midjourney Prompt Engineering for Photorealistic Portraits and Professional Product Shots
AI Prompts

Midjourney Prompt Engineering for Photorealistic Portraits and Professional Product Shots

M Morgan Johnson · · 3,249 views

Midjourney Prompt Engineering for Photorealistic Portraits and Professional Product Shots

Midjourney v7 (released early 2026) changed what's achievable with photorealistic portrait and product image generation. The gap between 'AI-generated' and 'photographed by a professional' is now genuinely small on the right subjects with the right prompts. I've been running Midjourney for client work for a year — brand campaigns, product catalog images, social media content — and the prompt patterns that reliably produce commercial-quality output are specific and non-intuitive if you're coming from DALL-E or Stable Diffusion.

Portrait Prompt Architecture: Lighting, Lens, and Reference Stacking

Portrait quality in Midjourney is determined by three factors in descending importance: lighting specification, virtual lens and camera settings, and style/reference layering. The prompt architecture that produces professional-quality portraits: '[Subject description with specific physical characteristics] — [lighting setup: key light position, light quality, fill ratio, background lighting] — [virtual camera: specific lens (85mm f/1.4 for classic portraits, 50mm f/2 for environmental), aperture feel, apparent focal distance] — [style reference: specific photographer's aesthetic, a publication style (Vogue editorial, National Geographic environmental), or era (1970s Kodachrome, contemporary clean digital)] — photorealistic, high detail, professional photography.' The lighting specification is the highest-leverage prompt element for portraits. 'Studio lighting' is vague. 'Rembrandt lighting from camera-left, soft fill from reflector right, rim light separating subject from dark background' produces something specific and professional-looking.

Midjourney v7 responds well to photographer name references: Annie Leibovitz (dramatic environmental portraits), Yousuf Karsh (classic studio gravitas), Peter Lindbergh (natural black and white), Chase Jarvis (natural light lifestyle). These aren't just style references — they activate complex combinations of lighting, framing, color grade, and subject relationship that would take many more words to describe explicitly.

Product Photography Prompts: Commercial-Quality Images Without a Studio

Product photography prompts need different architecture than portraits. The three elements that produce commercial-quality product images: background and surface specification, lighting setup, and context/lifestyle environment. Prompt pattern for a clean product shot: '[Product description with specific material, color, and finish] on [surface: white marble / brushed concrete / natural wood grain] — [background: pure white, gradient gray, environmental blur] — [lighting: even diffused studio light / dramatic side lighting / natural window light from [direction]] — [angle: directly above flat lay / 3/4 angle at [height] / directly front-facing] — product photography, commercial catalog quality, no shadows on background, hyperrealistic.' The 'no shadows on background' instruction is critical for catalog-style shots — AI tends to cast object shadows onto white backgrounds even when not instructed to. The lifestyle product shot uses a different structure: '[Product] in use in [specific realistic environment] — [natural light quality] — shallow depth of field, focus on product, editorial lifestyle photography, [specific publication style if relevant].'

For product consistency across multiple shots (you need the same product shown from 5 angles), use Midjourney's character/object reference feature (--cref or --sref flags on v7) to maintain visual consistency. Without this, the same product looks slightly different in every generation, which breaks product catalog usability.

Negative Prompts and Technical Parameters for Midjourney v7

Midjourney v7's parameter and negative prompt system is more nuanced than v6. Key parameters I use for professional output: --style raw (reduces artistic interpretation, closer to literal prompt execution — essential for product work), --stylize 0-1000 (0 = purely literal, 1000 = maximum creative interpretation; I use 200-400 for editorial work, 0-100 for product work), --ar for aspect ratio (16:9 for landscape editorial, 4:5 for Instagram, 1:1 for catalog), --q 2 (maximum quality — slower but noticeably sharper for print-scale output). Negative prompts via '--no' flag: for portraits, '--no extra fingers, lens distortion, plastic skin, oversaturation'; for product shots, '--no reflections in wrong places, blurry label text, bent surfaces, logo distortion.' The 'extra fingers' issue in portraits has dramatically improved in v7 but still occurs on some hand poses — the --no flag catches the remaining cases. Label text rendering is still Midjourney's weakest point for product work — readable text on product packaging requires specific approaches (generate without text, add text in Photoshop).

For client review rounds: use --seed [number] to reproduce a generation with modifications. The seed locks the composition and basic visual structure while allowing prompt modifications. Show a client three variants by using different seeds on the same prompt, then iterate on the approved seed.

This note was created with Penlify — a free, fast, beautiful note-taking app.

Start writing free Browse Explore