Skip to content

AI Features Overview

ArcBrush includes optional AI-powered nodes that wire into your graph like any other node. AI features require an internet connection and an account with credits.

Generate images from text prompts. Choose from four model tiers ranging from fast drafts to advanced quality. See AI Text to Image for full parameter details.

ModelTierSpeedBest For
FLUX Fast (Schnell)Fast~2–8sQuick iteration, concept exploration, lowest cost
FLUX.2 ProQuality~15–35sHigh-quality generation, professional output
Nano Banana 2Advanced~15–40sBest prompt understanding, optional web search, style transfer
FLUX.2 MaxPremium~20–45sHighest fidelity from Black Forest Labs

Edit existing images with text instructions. Feed in one or more reference images and describe the change you want. See AI Image Edit for full parameter details.

ModelTierSpeedBest For
FLUX.2 TurboFast~5–15sQuick instruction-based edits, lowest editing cost
FLUX.2 ProStandard~15–30sHigh-quality edits, professional workflows
FLUX.2 MaxQuality~20–40sHighest fidelity edits, complex changes
Nano Banana 2Advanced~15–40sRestyling, multi-image fusion, complex creative edits

AI background removal that produces a clean RGBA cutout with transparent background. Choose between two models at the same credit cost. See AI Remove Background for details.

ModelSpeedBest For
rembgVery fastReliable general-purpose removal, open-source
Bria RMBG 2.0FastHigher quality edges, especially on hair and fur (default)

4x super-resolution that adds realistic detail beyond what a simple resize can produce. See AI Upscale for details.

ModelBest For
v1General-purpose upscaling, good for all image types
v2Photographs and non-AI-generated images (recommended)

AI nodes don’t auto-evaluate like other nodes. You control when a generation happens by clicking the Run button (labeled “Generate”, “Edit”, or “Remove Background” depending on the node).

  1. Set your parameters (prompt, model, dimensions, etc.).
  2. The exact credit cost is shown in the node before you run.
  3. Click Run. The button changes to “Waiting in queue…” then “Generating…” while the request processes.
  4. The result appears in the Preview Panel and is automatically saved to disk.

The button is disabled while a generation is in progress. If you haven’t signed in yet, the button prompts you to sign in first.

Every generation is saved to a sidecar folder next to your .arcb project file. If your project is hero.arcb, generations are saved in hero.arcb_data/.

When a node has more than one saved generation, a history dropdown appears in the properties panel. You can select any previous generation to view it instantly. No API call is needed; it loads directly from disk.

  • Generations are numbered newest-first: “Generation 3 (latest)”, “Generation 2”, “Generation 1”.
  • Selecting an older generation doesn’t delete newer ones. The full history is preserved.
  • History is per-node, so each AI node in your graph maintains its own independent set of generations.

Before you click Run, each AI node displays the exact cost based on the selected model and output dimensions. Higher-resolution outputs and more capable models cost more credits. The exact cost is also logged after each generation.

Your credit balance is shown in the app’s status bar at the bottom, along with:

  • Session total: credits spent since launching the app
  • File total: credits spent on generations in the current project

Every generation (successful or failed) is recorded in a local log file. This log includes timestamps, prompts, model used, dimensions, credit cost, and duration. It’s useful for tracking your spending and reproducing past results.

If you change a prompt or input after generating, the node shows a precise staleness badge so you know exactly what diverged from the last successful run:

  • Input changed — an upstream image or connection changed.
  • Params changed — a parameter on the node itself changed.
  • Out of date — both the input and the parameters changed.

Click Run again to generate with the updated settings. Reverting a slider or undoing an upstream change clears the badge automatically. The badge state is saved with the project, so it survives close and reopen.

Most AI models accept an optional seed value. Using the same seed with the same parameters produces the same output, which is useful for reproducing results. Set the seed to -1 (the default) to let the model choose a random seed each time.

AI nodes are composable with every other node:

  • Feed images into AI Image Edit to restyle, combine references, or turn sketches into art.
  • Pipe AI output through Mask Refine to clean edges.
  • Connect to Palette Remap to generate color variants of AI-generated art.
  • Chain multiple AI nodes for complex multi-step workflows.

All 71 non-AI nodes (every filter, transform, mask tool, and export) work without an account, without signing in, without limits. AI features are entirely optional.

  1. Click Generate on any AI node.
  2. A browser window opens. Sign in with Google or email.
  3. Your account starts with 10 free credits.
  4. Purchase additional credit packs when needed.

See Credits & Pricing for details.