Restaurant Menu Photos: Lighting, AI Cleanup, Conversion

Why menu photos move revenue more than menu copy

Restaurant menus convert on two axes: visual appeal and clarity. For decades the dominant lever was copy. Descriptive item names, ingredient callouts, premium adjectives like 'hand-cut' or 'artisan' that menu engineering studies show lift order rates 15-30% over plain descriptions. That lever still works, but it's been overtaken by photos as the dominant conversion driver, mainly on delivery platforms where the customer is scrolling through twelve restaurants in 90 seconds and decides almost fully on which item images stop them.

Uber Eats and DoorDash both publish merchant-facing data showing that menu items with platform-quality photos see 1.5x to 3x the click-through rate of unphotographed items in the same listing. The lift is bigger for high-margin items (entrees, signature dishes) than for low-margin sides. The photo work disproportionately benefits items the customer was on the fence about ordering. A clear photo of a $24 entree tips the decision. The same customer would have ordered the $4 side of fries either way.

The other axis where photos compound is Google Business Profile. Listings with 10+ recent, high-quality photos rank higher in the local 3-pack and get more direction requests, calls, and website clicks than listings without. Google publishes this in the Business Profile help docs but doesn't quantify the lift. Restaurant marketing data from third-party sources suggests roughly a 40-70% lift in profile actions for restaurants that upload one new high-quality photo per week versus those that upload nothing. The same photos you took for DoorDash work here — repurpose, don't re-shoot.

• Platform photo quality drives 1.5x-3x click-through on DoorDash and Uber Eats.
• Higher lift on high-margin entrees than on low-margin sides.
• Google Business Profile photo activity drives ~40-70% lift in profile actions.
• One photo set serves multiple channels — repurpose instead of shooting per platform.

How to light a dish without studio equipment

Studio food photography uses two or three lights, a diffusion panel, and a reflector to control shadow and highlight. You don't need any of that. What works almost as well, and what every working restaurant has, is a window. Find the brightest window in the dining room. Usually north-facing or east-facing in the morning, west-facing in the afternoon — and set up a small table or a board on a chair beside it. The bigger the window, the softer and more flattering the light.

Turn off every other light source in the room. Overhead pendants, kitchen fluorescents, neon signs, and color-changing LED accent strips all introduce color casts that AI photo editing has to fight, and they create competing shadow directions that make the dish look chaotic. One light source produces one shadow direction, which is what reads as 'expert' to a diner scrolling at speed. The visual cue isn't expensive equipment; it's the simplicity of the lighting.

Direct sunlight is bad — it produces hard shadows and blown highlights on glossy surfaces (skin of a roast chicken, oil on a salad dressing). The fix is to wait for indirect light: late morning, mid-afternoon, overcast days, or shoot during the window when the sun is past the building. If direct sunlight is unavoidable, hang a sheer white curtain or a translucent white plastic shower curtain across the window. This diffuses the light at zero cost and instantly produces softer shadows. Restaurant operators don't need to know any of the technical terms. They need to know to use one big bright window with every other light off. To diffuse the window if the sun is hitting it directly.

• Find the brightest window — set up a table next to it.
• Turn off every other light. One light source = one shadow direction = professional look.
• Avoid direct sun; if unavoidable, diffuse with a sheer curtain or white plastic.
• Best windows: north/east in morning, west in afternoon, overcast days are forgiving.

Angle the camera based on the dish, not the photographer's habit

The most common amateur mistake in menu photography is shooting every dish at the same angle. Phones default to 45 degrees because that's the comfortable hand-hold position when standing next to a table. But the right angle depends fully on the shape of the plated dish. Flat-plated dishes — pizza, salad, grain bowls, flatbreads, sushi platters — are flattered by an overhead 90-degree shot that shows the full composition and uses the plate's circle as the frame. Layered or stacked dishes — burgers, club sandwiches, cakes, parfaits in shallow glassware — work at 45 degrees because that angle reveals the layers. Tall items in glassware — cocktails, milkshakes, tall coffees, parfaits in tall glasses — work straight-on at 0 degrees because that angle preserves the glassware's silhouette.

The cost of shooting at the wrong angle is real. A burger shot from directly overhead reads as 'patty on a bun-shape,' losing all the visual interest of the stacked composition that makes a burger sell. A salad shot at 45 degrees compresses into a chaotic green mass instead of presenting as a balanced bowl. Diners may not consciously think 'wrong angle,' but they scroll past the dish without ordering it. Is the only metric that matters.

The practical workflow: spend ten seconds before each shot deciding whether this dish is flat, layered, or tall. Switch angles. The shoot is slower by 20-30% but the keeper rate per dish doubles. Means fewer re-shoots and faster total time to a complete menu set. For restaurants photographing 30-50 menu items, that's the difference between two shoot days and one.

• Flat dishes: overhead 90° (pizza, salad, grain bowls, sushi platters).
• Layered dishes: 45° (burgers, sandwiches, cakes, parfaits in shallow glassware).
• Tall items: 0° straight-on (cocktails, milkshakes, parfaits in tall glasses).
• Wrong angle = scroll-past, even if every other variable is correct.

The AI editing workflow: three jobs only

The temptation with AI photo editing tools is to do too much. Enhance, sharpen, color-grade, denoise, upscale, AI-fill the background, and add fake steam. Every layer past the first one introduces 'fake food' tells that destroy conversion harder than a plain phone photo would. The disciplined workflow uses AI for three jobs only, in this order: clutter cleanup, boost, and background replacement (only when needed).

Clutter cleanup is the highest-ROI step. Use Magic Eraser to erase what shouldn't be in the frame: a crumb on the rim of the plate, the corner of a heat lamp that crept into the upper-right, a smudge on the surface, a partial hand from the person setting the plate down. Most amateur menu photos have one or two of these tells. Removing them with object-removal AI is 30-90 seconds per photo and lifts perceived quality greatly. This step has no downside — you're removing accidents, not adding fakery.

Boost is the second step, and the one most people overdo. One pass of AI boost to recover texture and color in the dish is correct. Two passes start producing the 'fake food' look. Over-saturated reds in roasted vegetables, neon greens in lettuce, glossy highlights on bread crust that look painted on. The diner's brain reads these as wrong even if they can't articulate why, and the conversion impact is negative. One boost pass is the rule. If the first pass didn't help, the source photo needed better lighting, not more boost.

Background replacement is the third step and is only sometimes needed. If the original background includes kitchen equipment, busy tile, or unwanted elements, swapping it for a clean staged surface (slate, wood, marble, coordinated linen) is worth the 60-90 seconds. If the original surface is already clean, skip this step. Replaced backgrounds, even good ones, are slightly less convincing than real surfaces shot in-camera, and adding the replacement when it isn't needed is a small unforced quality cost.

• Step 1: clutter cleanup with Magic Eraser. High ROI, no downside.
• Step 2: one AI enhancement pass. Two passes produces the 'fake food' look.
• Step 3: background replacement only if the original surface is cluttered.
• Discipline matters more than tool choice — over-editing kills conversion harder than under-editing.

Export to platform specs (DoorDash, Uber Eats, Google Business, in-house menu)

Different platforms display photos in different aspect ratios and different minimum resolutions. Submitting the same crop to every platform means each platform applies its own crop logic, and every crop loses something. The disciplined workflow exports a 4K master with the dish centered, then derives each platform's crop from that master with intentional composition for each aspect ratio.

DoorDash and Uber Eats both display menu item photos at roughly 4:3 or 16:9 aspect ratios with a 1400×1050 (DoorDash) or 1600×900 (Uber Eats) minimum. Composing the dish slightly off-center in the master gives you crop flexibility. When DoorDash applies a slight crop you don't lose part of the dish, and when Uber Eats applies a different crop you still have a balanced composition. Google Business Profile uses 1200×900 (4:3) and prefers higher resolution for the cover photo. In-house digital menu displays usually run 16:9 at 1920×1080 or 4K.

Practical tip: name the exported files with the platform in the filename — `signature-burger-doordash-1400x1050.webp`, `signature-burger-googlebusiness-1200x900.webp`. When the menu changes seasonally, replacement is fast because the filename tells you what each version is for. Save the 4K master to a separate folder per menu set so you can re-derive new platform crops as platforms change their specs (which happens every 12-24 months).

• Export a 4K master per dish, derive each platform crop from that master.
• DoorDash: 4:3 at 1400×1050 minimum.
• Uber Eats: 16:9 at 1600×900 minimum.
• Google Business: 4:3 at 1200×900, higher res preferred for cover.
• In-house digital menu: 16:9 at 1920×1080 or 4K.
• Name files by platform so seasonal swaps are fast.

Stay within FTC and state-level truth-in-advertising rules

Food photography is regulated under the FTC's general truth-in-advertising rules: the image must fairly represent what the customer receives. The famous violations of this rule are old chain-restaurant lawsuits where the photographed burger was twice the height of the actual served burger, or where the photographed salad had ingredients the actual salad didn't. The principle is consistent — what's depicted must be a reasonable representation of what arrives.

AI photo editing doesn't change this rule. Lighting and cleanup are fair: adjusting brightness, removing accidental clutter, color-correcting for a window's warm cast all preserve the dish as served. Adding garnish that isn't on the actual plate, faking steam to imply 'just out of the oven,' AI-generating extra ingredients (more pickle slices, a thicker patty), or expanding the portion size via outpainting are all in violation territory. The most common amateur trap is adding parsley or microgreens via AI fill 'to make the dish look more finished'. If the dish is served without that garnish, the photo can't include it.

State-level enforcement varies. California, New York, and a handful of other states actively pursue food advertising deception claims. Most states defer to FTC action. Class-action lawsuits from individual customers happen but are rare unless the gap between photo and served dish is large and the chain is national. For independent restaurants, the practical risk isn't a state AG investigation. It's the cumulative review damage from customers who feel misled, who write 'looked nothing like the photo' reviews, which drag down platform ranking faster than any amount of photo work can lift it.

• FTC rule: photo must fairly represent what's served. Lighting and cleanup are fair; adding ingredients is not.
• Common trap: AI-filling garnish onto a dish served without it. Don't do it.
• State enforcement varies — California and New York active, most states defer to FTC.
• Bigger practical risk for independents is review damage from misled customers.