Enhance the dish, never invent it
Change the light, background and steam, never the food: same portion, same ingredients, same plate. Compare every edit with your original photo before you post it.
Food photos · ChatGPT · October 2026
How do I make professional food photos in ChatGPT?
New light, clean background, the same food.
In ChatGPT, add the Creative Claw plugin, attach a phone photo of the dish and ask to change only the light, background and steam. In our test, FLUX Kontext Max did exactly that on three dishes for 16 credits each, while three other models changed the food (one dropped a taco). Use the retouch on your website and social. DoorDash asks for menu photos that are “clean and unedited,” so upload your real photo there.
Start with 100 free credits. No card required.
By Itay Dressler, founder of Creative Claw · LinkedIn · Updated
Key takeaways
- FLUX Kontext Max kept both test dishes unchanged; the other three models each changed the food.
- Name every item that must stay: count, garnish, cup, plate.
- DoorDash: “clean and unedited”, the actual item, landscape 16:9.
- Google Business Profile: “no significant alterations”.
- A retouch is 16 credits; photographers start around $42 a dish.
Bring
A phone photo of the dish, whole plate in frame
Create
A retouched photo with the food unchanged
Made with Creative Claw
What can you make with Creative Claw in ChatGPT?
Before and after
Same dish, better light (and one that went wrong)
The first three are FLUX Kontext Max retouches with the food unchanged. The last is another model’s edit of the same birria photo: it dropped a taco.
Original
AI retouch 
Original
AI retouch 
Original
AI retouch Original
Changed the dish Finished example
The photos in context: one restaurant’s whole kit.
We invented a restaurant, Brasa Azul, an Oaxacan-Californian grill in San Diego’s North Park, and started from three ordinary phone photos taken at the pass. FLUX Kontext Max retouched the light and background without touching the food, Gemini Omni Flash shot three reels, ElevenLabs v4 voiced the promo in English and Spanish, Lyria 3.5 sang a bilingual jingle, Nano Banana 2 lettered the chalkboard, and an HTML render made a printable menu with exact prices.
Voiceover script
“Slow-cooked mole. Mole a fuego lento. Blue corn, pressed by hand. Maíz azul, hecho a mano. Mezcal, with a little fire. Mezcal, con un poco de fuego. Brasa Azul. Now open in North Park. Ya abierto en North Park.”
| Step | Tool | Credits |
|---|---|---|
| Three dish retouches that keep the food unchanged | FLUX Kontext Max | 48 |
| Three reel stills | Nano Banana 2, 2K | 72 |
| Three 6-second reels, 1080p | Gemini Omni Flash | 414 |
| Storefront photo with the neon sign | Nano Banana 2, 2K | 24 |
| Chalkboard specials board | Nano Banana 2, 2K | 24 |
| Printable menu with exact prices | HTML render | 1 |
| Bilingual jingle | Lyria 3.5 | 16 |
| English and Spanish voiceovers | ElevenLabs v4 | 8 |
| Lyric check on the jingle | Transcribe | 2 |
| Promo and menu board videos | HyperFrames (render_html_video) | 11 |
| Total for the finished example | 620 ($6.20) | |
The restaurant, dishes and phone photos are fictional and were generated for this page; with your own phone photos, skip nothing but the photos themselves. We also spent about 260 more credits testing which model keeps a dish unchanged (results on the food photos page).
-
Retouch · FLUX Kontext Max -
Reel still · Nano Banana 2 -
Specials board · Nano Banana 2 -
Storefront · Nano Banana 2
The workflow
How do I retouch a food photo in ChatGPT?
-
Add Creative Claw and attach the photo
Install Creative Claw from ChatGPT’s plugin directory and sign in. New accounts get 100 free credits, no card. Attach a photo with the whole plate in frame.
-
List what must stay
Name the count, garnish, sauces, cup and plate, and the few things to change: light, background, steam.
-
Retouch with FLUX Kontext Max
Ask for “retouch only, not a redesign” with FLUX Kontext Max; it is 16 credits.
-
Compare before you post
Check the edit against your photo item by item, then use it on your website and social, not on DoorDash.
Pricing
How much does it cost?
Creative Claw is pay as you go: $10 buys 1,000 credits, new accounts get 100 free credits, and there is no subscription. Typical costs for this workflow:
| What you generate | Credits | Cost |
|---|---|---|
| Dish retouch (FLUX Kontext Max) | 16 | $0.16 |
| Dish retouch (Nano Banana 2, 2K) | 24 | $0.24 |
| Lifestyle scene (Nano Banana 2, 2K) | 24 | $0.24 |
| Our four-model test on two dishes, one try each | 156 | $1.56 |
Prices charged or quoted by Creative Claw on October 25, 2026. See all model prices, and ask for an estimate in ChatGPT before you generate.
Alternatives
Which AI model keeps the dish the same?
The tools below are the ones AI assistants most often recommend for this job. We checked each vendor’s own pages on October 25, 2026. “None found” means we could not find an official app in the ChatGPT directory or on the vendor’s site. Creative Claw publishes this page.
| Model | Birria tacos | Mole negro | Credits |
|---|---|---|---|
| FLUX Kontext Max | Same three tacos, cheese, cup and limes | Same chicken, sauce and rice | 16 |
| Nano Banana 2 | Close, but the plate shrank and the cheese faded | Moulded the rice; a retry added peas | 24 |
| GPT Image 2.5 Sunburst | Re-plated the tacos open-faced, no cheese | Changed the chicken and rice | About 8 |
| Nano Banana Pro | Only two tacos, no cheese | Moulded the rice into a dome | 30 |
Sources: DoorDash: common reasons for rejected menu photos · Google Business Profile photo guidelines · Uber Eats AI photo enhancement (The Decoder). Prices are in USD before tax and change often.
Use cases
Where can you use AI-retouched food photos?
| Where | Rule | What to do |
|---|---|---|
| DoorDash menu | Photos must be “clean and unedited” and show “the actual item served, not an idealized or generic version”; landscape 16:9 only; photoshopped backgrounds are rejected | Upload your real photo, shot in landscape on a plain surface |
| Uber Eats menu | Uber itself uses AI to improve low-quality merchant photos (lighting, resolution, framing, plating) | Check how your dishes appear in the app; keep a real photo on file |
| Google Business Profile | Photos should have “no significant alterations” and “should represent reality” | Post real photos; light and background clean-up only |
| Your website and Instagram | No platform photo rule, but deceptive food ads are still deceptive | Retouch light and background; never change the food |
| Reels and TikToks | TikTok and Meta ask creators to label realistic AI-generated video | Use the platform’s AI label on AI reels |
Before you render
What should I check before rendering?
Delivery apps want real photos
DoorDash asks for menu photos that are “clean and unedited” and show “the actual item served, not an idealized or generic version.” Upload your real photo there.
Check the count and garnish
Count the tacos, the slices, the scoops; check garnish placement and the cup or plate. Our failed edits changed exactly these.
Keep a real photo on file
Keep the original for delivery apps and Google, and in case a platform asks.
Check current pricing, read short answers in the help center, inspect the tool reference, or browse published media examples. Examples illustrate creative approaches; they are not proof that this exact starter prompt was run.
Practical answers
Frequently asked questions
Food photos in ChatGPT
Can ChatGPT retouch my food photos?
Yes, with Creative Claw. Attach the phone photo and ask to change only the light, background and steam. FLUX Kontext Max did that on our three dishes without touching the food, for 16 credits each.
Which model keeps the dish the same?
In our test, FLUX Kontext Max. With the same strict prompt, Nano Banana Pro dropped a taco and the cheese, GPT Image 2.5 Sunburst re-plated the tacos and changed the chicken, and Nano Banana 2 moulded the rice, then added peas on a retry.
What should the prompt say?
Name every item that must stay (count, garnish, cup, plate) and the few things to change (light, background, steam). “Retouch only, not a redesign” helps. Then compare the result with your photo, item by item.
Can I use the retouched photo on DoorDash?
No. DoorDash wants menu photos “clean and unedited” and rejects photoshopped backgrounds. Use the retouch on your website and social, and shoot the DoorDash photo in landscape on a plain surface.
What makes a good phone photo to start from?
The whole plate in frame, centered, in daylight if you can, landscape for delivery apps. DoorDash asks for at least 80% of the item visible and at least 1400 × 800 pixels.
Can it make lifestyle scenes with my dish?
Yes, for mood posts: a table set for two, a patio at dusk, a hand pouring sauce. Keep the dish itself recognisably yours, and do not use those scenes as menu photos.
Can AI make a portion look bigger or add toppings?
It can, and you should not let it. A bigger portion or extra toppings in a photo is deceptive advertising. Tell the model what must stay the same, and reject any edit that changes the food.
Do I have to tell customers a photo was retouched?
Light and background clean-up is normal food photography. Delivery apps and Google set their own limits (DoorDash wants unedited menu photos). If an image is a generated scene, not your real dish, do not present it as a menu photo.
Cost
How much does an AI food photo cost?
A FLUX Kontext Max retouch is 16 credits ($0.16); a Nano Banana 2 scene at 2K is 24. Food photographers start around $42 a dish with a $499 minimum.
Delivery apps, ads and the rules
Can I use AI food photos on DoorDash?
Not edited ones. DoorDash’s help center says menu photos must be “clean and unedited,” must show “the actual item served, not an idealized or generic version,” must be landscape 16:9, and that photoshopping a background onto a photo is cause for rejection. Upload your real photo, well shot.
Does Uber Eats allow AI-enhanced photos?
Uber Eats uses AI itself: since July 2025 it improves low-quality merchant photos (lighting, resolution, framing and plating), and press reports say it can change plates and backgrounds. Check how your dishes look in the app.
What does Google Business Profile allow?
Google’s photo guidelines ask for “no significant alterations or excessive use of filters” and say the image “should represent reality.” Post real photos; cleaning up light and background is the most to do.
Can I show a dish that is not on my menu, or a bigger portion?
No. Showing food you do not serve, or more than you serve, is deceptive advertising. Generate scenes and mood freely, but keep every plate in an ad or menu photo true to what customers get.
Can I post AI-generated reviews or diner testimonials?
No. The FTC rule in force since October 21, 2024 bans fake reviews and testimonials, including AI-generated ones, with civil penalties for knowing violations. Quote real reviews you have permission to use.
Do I need to label AI reels on Instagram and TikTok?
TikTok requires labels on realistic AI-generated content, and Meta asks creators to disclose photorealistic AI video. Turn on the platform’s AI label for reels made with AI.
Setup and pricing
How do I add Creative Claw to ChatGPT?
Open chatgpt.com/plugins/creativeclaw and select Install plugin, or search for Creative Claw in ChatGPT’s plugin directory, then sign in. New accounts get 100 free credits, no card.
Do I need a subscription?
No. Creative Claw is pay as you go: $10 buys 1,000 credits that never expire, and new accounts get 100 free credits. You pay for the posts you make, not per month.
Can I use my own photos and menu?
Yes, and you should. Attach your phone photos and paste your menu in ChatGPT; the retouches, menus and boards are built from your real dishes and prices.
Does Creative Claw post to Instagram, Google or DoorDash?
No. It makes the files (photos, reels, menus, boards); you upload them to your accounts, your menu manager or your scheduler.
Start creating
Install Creative Claw in ChatGPT.
Connect once, get 100 free credits with no card required, and use the starter prompt above for your first project.
