Most articles about the cost of an AI photo editor app come from agencies, and their numbers run from $20,000 to over $500,000. That range is too wide to plan with, and it leaves out the part that decides whether the app survives: every AI edit costs money, every time a user taps the button.
This guide splits the cost into the two parts that matter: what it costs to build the app, and what it costs to run it per image. It is based on Evoke AI, an AI photo app I co-built with one other developer. It is live on the App Store and Google Play, with 4,000+ active members. Prices are from the providers’ official pages in November 2026; they change often, so check the linked pages before you budget.
Quick answer: a focused AI photo editor MVP (sign-in, 2–3 AI features, credits and a paywall) usually takes about 600–1,000 hours to build with Flutter, roughly $15,000–$40,000 with an experienced South Asian developer and much more with a US agency. Running it costs about $0.03–$0.13 per generated image with Google’s current image models, plus store fees of 15% on most small apps’ sales and a server from about €5 a month.
What we actually built
Evoke AI is an app and a website. The app has an AI chatbot, prompt-to-image generation, AI photo editing by prompt, a moments feed and a set of file converters. The website has more than 60 AI tools and converters. Two of us built it in about six months.
The parts that took real engineering were not the AI calls themselves:
- In-app purchases for credits, end to end.
- Ads, including ads that unlock a prompt.
- A server-side quota: every new account gets 5 free prompts, then buys credits or watches an ad per prompt.
- Our own backend on a Contabo server, so the app never holds an AI API key.
That last point is not optional. If the key is inside the app, anyone can extract it and run up your bill. How to set this up: calling an AI API from Flutter without exposing your key.
Build cost, feature by feature
These are planning estimates for a Flutter app on iOS and Android with one experienced developer, in the same ranges as my Flutter app development cost guide. They are not Evoke’s invoice.
| Part | Hours (estimate) | Why it takes that long |
|---|---|---|
| App shell: sign-in, profile, settings, gallery | 120–200 | Standard screens, plus account deletion both stores require |
| 2–3 AI features (generate, edit by prompt, remove background) | 120–220 | Upload, progress, errors, retries, saving and sharing results |
| Credits and paywall (in-app purchases) | 60–120 | Store products, receipts, restoring purchases, edge cases |
| Rewarded ads | 30–60 | Ad SDK, consent, granting the reward on the server |
| Backend: AI proxy, quota, credits ledger | 120–220 | The part that protects your API bill |
| Admin basics and analytics | 40–80 | Seeing usage and cost per user |
| QA, store review and release | 110–180 | Testing AI results, content rules, store listing and review |
| Total | 600–1,080 |
At the hourly bands from the cost guide, that is roughly $15,000–$43,000 at South Asia rates ($25–$40/h), $21,000–$81,000 in Eastern Europe ($35–$75/h) and $60,000–$162,000 with a US agency ($100–$150/h, assumption). A website with dozens of extra tools, like Evoke’s, is a separate project on top.
To estimate your own scope, use the free app cost calculator and tick the AI and payments features.
Running cost: what each AI edit costs
This is the cost most founders miss. You pay the AI provider for every image, whether or not that user ever pays you.
| Model (provider) | Price per image | Notes |
|---|---|---|
| Gemini 3.1 Flash Lite Image (Google) | about $0.034 at 1K | Cheapest current Google option |
| Gemini 3.1 Flash Image (Google) | about $0.067 at 1K | |
| Gemini 3 Pro Image (Google) | $0.134 at 1K–2K, $0.24 at 4K | Highest quality, highest cost |
| OpenAI image models | billed by tokens, from $8 (gpt-image-1-mini) to $32 per 1M image output tokens | Per-image cost depends on size and quality; use OpenAI’s calculator |
| remove.bg (background removal) | about $0.07–$0.23 | Falls with volume on its subscription plans |
Sources: Gemini API pricing, OpenAI API pricing, remove.bg pricing. Both Google and OpenAI also offer a Batch option at half price, which only suits jobs that don’t need an instant result.
A worked example
Say 1,000 people install your app in a month and each uses the 5 free prompts:
- 5,000 images × $0.034 = $170 with the cheapest Gemini model.
- 5,000 images × $0.134 = $670 with Gemini 3 Pro Image.
That is before a single sale. If the free tier is unlimited instead of 5 prompts, the bill has no ceiling. This is why the free quota has to be enforced on your server, not in the app, where a modified copy of the app could skip it.
Store fees and servers
Store commission. Apple’s Small Business Program takes 15% on paid apps and in-app purchases for developers with up to $1 million in proceeds the year before. Google Play’s service fees are 15% on the first $1 million a year in most markets. In the US, EEA, UK, Japan and Australia, Google is moving to a new structure: 10% plus a 5% billing fee for subscriptions, and 20% plus 5% for other purchases on new installs. Budget for the commission when you price credits.
Servers. The backend that holds your keys and counts credits is light. A VPS like Contabo’s Cloud VPS starts from about €4.40–€5.50 a month (4 vCPU, 8 GB RAM), which is enough to start. Add a database and file storage for user images; Firebase or Supabase free tiers often cover an MVP.
Make the AI pay for itself
The model that worked for us in Evoke AI:
- A small free quota (5 prompts) so people can try the app.
- Then credits or an ad. Users buy credits, or watch a rewarded ad for one more prompt. Every prompt is either paid for or covered by ad revenue.
- Everything checked on the server: quota, credits and ad rewards.
Price credits from your real cost: AI cost per image, plus the store’s commission, plus a margin. If a credit pack sells for $4.99 and the store keeps 15%, you receive about $4.24. At $0.034 per image, even 50 credits cost you only $1.70 in AI. At $0.134 per image, the same 50 credits cost $6.70, more than you receive. Choose the model before you set the price, not after.
Mistakes that make AI apps expensive
- An API key inside the app. It will be extracted. Keep keys on your server.
- No per-user limits. One script or one heavy user can burn a month’s budget in a day.
- The most expensive model for everything. Use the cheap model for previews and drafts, the premium one only for final exports users pay for.
- No cost tracking. Log the AI cost of every request per user from day one, so you can see if a feature loses money.
- Ignoring content rules. If users can share generated images with each other, Apple’s guideline 1.2 on user-generated content applies: filtering, reporting and blocking. Plan for it before review, not after a rejection.
Questions
How much does it cost to build an app like Remini or Lensa?
A focused version with 2–3 AI features, credits and a paywall is usually 600–1,000 hours of work with Flutter, about $15,000–$40,000 with an experienced South Asian developer. Matching every feature of a large app costs far more. The bigger ongoing cost is the AI itself, at roughly $0.03–$0.13 per image with current Google models.
Is it cheaper to train my own AI model?
Almost never for a first version. Training and hosting your own image model means GPU servers and machine learning work that cost far more than API fees at MVP scale. Start with a provider’s API, measure your real usage, and only consider your own model when the API bill is large and predictable.
Can I build an AI photo editor with Flutter?
Yes. Flutter handles the app on iOS and Android from one codebase, and the AI runs on a backend you control. Evoke AI is built this way. On-device features such as face detection or simple filters can run in the app itself without any API cost.
How do AI photo apps make money?
Usually with a small free quota, then credit packs or a subscription, and often rewarded ads for users who don’t pay. The key is that every AI request is either paid for or covered by ad revenue, and that the limits are enforced on the server.
Planning an AI photo app?
I’m Muhammad Naeem, a Flutter developer in Lahore with 10+ apps live, including Evoke AI. If you are planning an AI photo or creative app, tell me about your idea and I’ll reply with a scope, the AI cost per user and a price range. See also AI app development.
Need a Flutter app built? Tell me about your project, or see case studies of apps I’ve worked on.
