Flutter AI Image Generator App Boilerplate

An AI image app lives and dies on cost control

Image generation is expensive per call, so the make-or-break of an AI image app is not the pretty grid — it is metering. A user must be stopped at zero credits, a paying user's allowance must refill, and your OpenAI key can never leak, because a leaked key on an image endpoint is a runaway bill. The Flutter Kit's AI image base is built around exactly that. Generations route through a secure Flask proxy so the key stays server-side; a RevenueCat paywall sells the plan; and a credits Cubit decrements on each generation and blocks when empty. All of it lives in BLoC/Cubit with get_it, so the spend-control logic is testable and outside your widgets.

• DALL·E calls proxied through Flask — API key never in the bundle
• Credits Cubit decrements per generation, blocks at zero
• RevenueCat paywall ties a plan to a credit allowance
• BLoC/Cubit + get_it keeps generation and metering testable

Why this is distinct from a chat AI app

It is worth being explicit, because the kit also supports streaming chat: an image app is a different shape. Chat streams tokens and the cost scales with conversation length; image generation returns a binary image, costs are chunky and per-request, and the UX is prompt → wait → reveal → save/share. This base targets the image flow specifically: a generation request, a loading state worth designing for (these calls take seconds), a result you display and persist, and a gallery of past creations. If you want a conversational assistant instead, the chat module is the better fit — this page is about turning prompts into pictures.

• Prompt → generate → reveal → save/share flow, not a chat thread
• Designed-for loading state — generations take real seconds
• Generated images persisted to a gallery the user owns
• Style/size parameters passed through to DALL·E via the proxy

Generation and credits, concretely

A generation goes through a repository that calls your Flask proxy; the proxy holds the OpenAI key and returns the image. On the client, a Cubit checks credits before firing, decrements on success, and exposes a clear generating/done/error state so the button can't double-spend. Credits map to a RevenueCat entitlement — a subscriber's plan tops up their balance — and the paywall appears when they run dry. Everything themes from Material 3 design tokens, so your generator screen matches your brand from one file.

class ImageGenCubit extends Cubit<ImageGenState> {
  ImageGenCubit(this._gen, this._credits) : super(const GenIdle());
  final ImageRepository _gen;     // calls Flask proxy, key stays server-side
  final CreditsCubit _credits;

  Future<void> generate(String prompt) async {
    if (!_credits.canSpend(1)) { emit(const OutOfCredits()); return; }
    emit(const Generating());
    try {
      final url = await _gen.createImage(prompt: prompt, size: '1024x1024');
      _credits.spend(1);
      emit(GenSuccess(url));
    } catch (e) {
      emit(GenError(e.toString()));
    }
  }
}

When a different starting point is the honest choice

If your edge is a custom or fine-tuned image model, your own GPU inference, or heavy server-side pipelines (upscaling, in-painting, batch jobs), the value is in your backend and ML ops — the app is the thin client, and this kit's DALL·E-via-proxy default is a starting point you will heavily extend. If you want a conversational assistant rather than image generation, the chat-focused AI boilerplate fits better. And if you have no Dart experience and only need a single-prompt toy, a no-code tool may be faster for a throwaway v1. The Flutter Kit's AI image base is the right call when you want a code-first image-generation app you fully own, with the proxy, paywall, and credits metering already wired so you can focus on prompts, styles, and the gallery.

The Flutter Kit vs building an AI image app yourself

Frequently Asked Questions