Midjourney Prompts for Product Photos That Convert

The short version: Midjourney can produce product photos good enough for social ads and website banners, but the prompt structure matters more than the words you choose, and there's one thing it still can't do reliably that nobody tells you upfront. Use a formula (subject, surface, lighting, lens, background, style tags) and expect to generate 15 to 20 images to get 2 or 3 usable ones.

Why product photos are a different job from portraits or landscapes

Most Midjourney guides are written for people making fantasy art or portraits. Product photography has a different problem: the product has to look like itself. If you're selling a ceramic mug with a specific glaze colour, Midjourney doesn't know your glaze colour. It'll invent something close and confident, which is the opposite of what you need when a customer is comparing the photo to the actual item they'll receive.

This is the bit most "best Midjourney prompts" listicles skip over: AI-generated product shots are brilliant for mood, backgrounds, and lifestyle context, and weak at rendering your exact product with accuracy. If you need the product itself to be pixel-perfect, you're better off photographing the real item and using Midjourney (or a similar tool) to generate the background and styling around it, then compositing. Anyone telling you to just type a prompt and get a flawless shot of your actual product is selling you something.

The prompt formula that works

Strip away the fluff and a solid product prompt has six parts, in this rough order:

• Subject - what the product is, described plainly (not "stunning", just facts: material, shape, colour)
• Surface - what it's sitting on (marble, linen, concrete, water)
• Lighting - soft studio light, golden hour, hard directional light, softbox
• Lens/camera style - macro, 85mm, shallow depth of field, product photography style
• Background - plain, gradient, lifestyle scene, out of focus
• Style tags and parameters - --ar for aspect ratio, --stylize, --v for version

A working example: "amber glass skincare bottle with black pump lid, placed on wet grey slate, soft diffused studio lighting from the left, 85mm lens, shallow depth of field, white background, product photography, clean commercial style --ar 4:5 --stylize 250 --v 6"

Run that and you'll get a bottle shape in the right ballpark, nice light, good composition. What you won't get is your exact label, your exact pump mechanism, or your exact amber tone every time. That's the trade you're making.

A worked scenario: candle brand needing Instagram and website shots

Say you run a small candle business selling 200ml soy candles in matte black vessels, and you need six product images for a seasonal campaign, three for Instagram (square), three for the website (portrait). Here's how that might go in practice.

1. Start with a base prompt for the "hero" shot: "matte black ceramic candle vessel with visible wick, lit flame, placed on dark walnut wood surface, warm amber lighting from a single candle flame, soft shadows, cinematic product photography, moody autumn tone --ar 4:5 --v 6"
2. Generate a batch of 4, pick the composition you like best, then use that as a reference image with the --iw (image weight) parameter to keep generating variations with consistent framing.
3. For the Instagram square crop, add "--ar 1:1" and drop the depth of field slightly so the vessel reads clearly on a small phone screen.
4. For a flat lay version, change the surface and angle: "overhead flat lay, matte black candle vessel surrounded by dried orange slices and cinnamon sticks, linen fabric background, soft natural window light --ar 1:1 --v 6"
5. Out of roughly 20 total generations across these variations, expect to keep 4 to 6. That ratio is normal, not a sign you're prompting badly.
6. For the final step, if the vessel shape or label in the AI image doesn't match the real product closely enough, swap in a real photo of the vessel and only use the AI backgrounds and lighting as a separate layer, blended in Photoshop or Canva.

That last step is the one people skip, and it's why so many AI product shots on small business Instagram pages look slightly "off" to anyone who's seen the real product. The fix isn't a better prompt, it's accepting Midjourney's limits and using it for what it's good at.

Lighting and surface words that do the most work

Certain words consistently push Midjourney toward commercial-looking results rather than arty renders:

• "Softbox lighting" or "diffused studio light" instead of just "bright" - gives you the flat, even light of a real product shoot
• "Catalogue style" or "e-commerce photography" - nudges the composition towards centred, clean, sellable
• "Negative space" - leaves room for text overlay if you're building an ad
• "35mm film grain" or "Kodak Portra tone" - adds warmth without tipping into fantasy rendering
• Naming a surface specifically (travertine, brushed steel, raw oak) rather than "nice background" - vague words get vague, often wrong, results

If you want more control over the actual camera feel, it's worth spending time with camera and lens prompts built for AI tools, because the difference between "product shot" and "looks like a stock photo from 2015" often comes down to lens language, not subject description.

Parameters worth knowing for product work specifically

• --ar 4:5 for most e-commerce and Pinterest-style listings
• --ar 1:1 for Instagram grid consistency
• --ar 16:9 for website banners and hero sections
• --stylize 100 to 250 keeps results closer to photographic realism; anything above 500 starts drifting towards illustration
• --chaos 0 to 10 keeps your batch consistent, which matters if you need a matching set for one campaign
• --no plus unwanted elements (hands, text, reflections, people) stops Midjourney adding things you didn't ask for, which it does often

On that last point, a full negative prompt list for AI images saves a huge amount of regeneration time, because Midjourney has strong habits (random hands, odd text, extra props) that are easier to block than to prompt around.

Where Midjourney loses to the alternatives

If your product has fine text, a logo, or a label that needs to be legible, Midjourney is the wrong tool full stop. It still mangles text more often than not, even on v6. For anything involving editing a real photo of your actual product (changing the background behind a real photographed item, removing a shadow, swapping a surface under a real object), tools built for photo editing rather than generation tend to do better. It's worth comparing approaches rather than assuming one tool does everything; our breakdown of Nano Banana versus Midjourney covers which jobs suit which tool, and Gemini prompts for image editing is the better starting point specifically for editing real product photos rather than generating new ones from scratch.

A realistic cost and time comparison

A single professional product photography session in the UK typically runs £150 to £400 for a handful of items with basic retouching, more for lifestyle sets with props and models. Midjourney costs £8 to £24 a month depending on plan, and an afternoon spent prompting can produce dozens of background and mood options. The honest trade-off: cheap and fast for backgrounds, moods, and concept testing; not a replacement for accurate photography of the product itself. Plenty of small businesses use both, real photography for the product, AI for seasonal backgrounds and social content, and get the best of each without pretending the AI route is a full substitute.

If you're weighing whether this is something to learn yourself or hand to someone, that's the same calculation worth running before hiring any AI consultant for a small business, because the tool cost is never the real cost. The real cost is the hours spent regenerating batches of 20 to find 4 usable shots, which adds up fast if your time is billed at anything above minimum wage.

A simple checklist before you start prompting

• Decide whether the product itself needs to be exact (use real photos) or whether mood and background matter more (Midjourney works well here)
• Write your subject description in plain, factual terms, not adjectives
• Pick one lighting style and one surface and stick to them across a whole batch for consistency
• Set your aspect ratio before generating, not after cropping
• Keep a negative prompt list handy so you're not fixing the same issues every batch
• Budget for 15 to 20 generations per final image you need

Frequently asked questions

Can Midjourney generate accurate photos of my exact product?

No, not reliably. Midjourney invents details rather than copying a reference exactly, so logos, labels, and precise colours usually come out wrong; it's better suited to backgrounds, moods, and lifestyle context around a real product shot.

How many images should I expect to generate for one usable product photo?

Plan for 15 to 20 generations to land on 2 to 3 keepers. That ratio is normal and not a sign of weak prompting, it's just how the tool works.

What aspect ratio should I use for e-commerce product photos?

Use --ar 4:5 for most e-commerce listings and Pinterest-style formats, --ar 1:1 for Instagram grids, and --ar 16:9 for website banners and hero images.

Is Midjourney cheaper than hiring a product photographer?

Per month it's far cheaper (£8 to £24 versus £150 to £400 for a basic photography session), but it doesn't replace accurate photography of the actual product, only the background, mood, and concept work around it.