Monet Lumiere app icon

Build a scene.
Keep every asset consistent.

Monet Lumiere is a macOS studio for AI image generators: Google Gemini, OpenAI GPT-Image, and ByteDance Seedream. Compose a picture out of layers, then regenerate any one of them on its own. Each layer remembers the prompt, model, and reference images that made it, so the asset comes back looking like itself.

macOS 13+

Monet Lumiere's Generate view — a prompt for a racing car photograph with the Nanobanana Pro model selected, above five generated images including a Porsche GT3 RS, an oil can, and a garage interior

Everything a finished scene needs.

Scenes, not one-off images

Build a picture out of layers: images from your library, results from a flow, or a fresh prompt. Arrange them on a canvas with alignment guides and export the finished scene as a PNG.

Regenerate one asset, not the scene

Each layer stores the prompt, model, and reference images that produced it. Not happy with the car? Regenerate that layer alone. The garage behind it stays exactly as you left it.

Three AI image generators

Google Gemini, OpenAI GPT-Image, and ByteDance Seedream through fal.ai, at up to 4K. Every model keeps its own controls for resolution, aspect ratio, and reference images.

A library that remembers

Every image lands in a fast local library with the prompt and settings that made it, and prompts worth keeping get saved with their model settings. No more folders of unnamed files.

Flows for repeat work

Chain the steps you always run, like generate then remove the background, into one action. A scene layer can point at a flow, so re-running it updates the layer in place.

Chat to iterate

Talk an image closer to what you meant, with the full history of how it got there. Anything you like goes into the library and on to a scene.

Change one thing, not everything.

Ask any AI image generator for the same subject twice and you get two different subjects. That is the problem scenes solve. A layer in Monet is not a flat picture, it is the recipe that made one: the prompt, the model, the settings, and the reference images you fed it. Select a layer, press Regenerate, and only that asset is redrawn, anchored to the same references so it still looks like itself.

  • Every layer keeps its own prompt, model & references
  • Regenerate a single layer in place
  • Alignment guides and one-click PNG export
Monet Lumiere's scene editor — four generated sports cars composed to scale on a dark canvas with measurement guides and a layers panel

Nothing you make gets lost.

Generating the image is the easy part. Remembering which of five takes was the good one, and what prompt made it, is where other tools give up. Monet keeps every result on your Mac, tagged by the model and the feature that produced it, and ready to drop straight into a scene.

  • Every image stored with the prompt that made it
  • Filter by model, flow, or chat origin
  • Saved prompts re-run in one click
Monet Lumiere's image library — a grid of generated car and prop images, each tagged with its source model

Pipelines, not busywork.

Some images always need the same follow-up: a background removed, a variant regenerated. A flow chains those steps once and re-runs them as a single action. Point a scene layer at a flow and running it again updates that layer, so a whole composition can be rebuilt from its parts.

  • Multi-step pipelines in one action
  • Built-in background removal
  • Flow output feeds your library and scenes
Monet Lumiere's Flows view — a list of saved two-step workflows, each pairing an image generation step with background removal

From API key to finished scene.

There is no account and no sign-up. Add a key for the image generator you want to use, then work down the sidebar.

  1. Connect an image generator. In Settings → API Keys, add at least one: Google for Gemini, OpenAI for GPT-Image, fal.ai for Seedream. Anthropic is optional, for Chat.
  2. Make your first image. Write a prompt in Generate, pick a model and its settings, and attach reference images if you have them. Jobs run in the background and every result is saved to your library.
  3. Build the scene. In Scenes, add a layer from your library, from a flow, or from a new prompt. Drag the layers into place; guides snap them to each other.
  4. Regenerate what needs it. Select a layer, tweak its prompt or references, and choose Regenerate. Only that layer is redrawn. Export the finished scene as a PNG when it looks right.

Your work stays on your Mac. Keys, projects, prompts, scenes, and the image library are stored locally. Only a request’s prompt, chat context, and attachments go to the image generator you picked.

Read the full documentation →

Stop starting over.

One purchase, every image generator you hold a key for. If you are tired of losing a good asset the moment you regenerate it, give Monet a try.