Guide

Getting started

Get from a blank canvas to your first generated image. This chapter is the bare minimum — a handful of nodes and one Output. Once the basics click, grow more complex trees to take more control of your scenes; the graph scales with you.

Credits and keys

Sign in at flow.w33s3.com. Your membership includes a monthly allowance of credits, and there are two ways to power generation — use either, or both:

Your credits

Generation draws on your balance by model: fast image models run 1–2 credits per image, the premium ones 3–5; video is billed per second of output. Each Output node shows a generation's cost right on its Generate button before you run it, and your balance lives in the top banner of your dashboard, next to Buy Credits. If you burn through a month's allowance mid-project, top up with a credit pack from there — nothing is lost, your graph just waits.

Bring your own keys

Click Settings in the left sidebar and add API keys for the providers you use:

Provider Where to get a key
Google Gemini Google AI Studio
fal.ai fal.ai Dashboard
OpenAI OpenAI Platform
Stability AI Stability AI Platform

When a model has your own key configured, FlowBoard calls the provider directly and no credits are consumed — you're billed by the provider instead. Models without a key fall back to your credits automatically.

A good starting point for new users: Gemini Flash with a free key from Google AI Studio — fast, capable, and costs nothing while you learn.

The Settings panel's API Keys section: Gemini, fal.ai, and OpenAI showing Connected with redacted keys, Stability AI empty with a paste-your-key field, plus generation defaults below
Settings → API Keys — connected providers route direct; empty ones fall back to credits.
An Output node showing 'Using: Gemini 3 Pro' above its generated image and a Generate button labeled 'Generate (5 cr)' — the per-generation credit cost shown before you run it
The cost is on the button — this Output runs Gemini 3 Pro at 5 credits per generation.

One tip that applies either way: the Mock model costs nothing at all. It returns placeholder images instantly — use it while you're learning the graph, and switch to a real model when the wiring is right.

Your first character

A character portrait is the perfect first project: it teaches the graph, and the result becomes a reference asset you'll reuse in every scene that character appears in. This walkthrough builds Ray — a club manager from a 1990 Atlanta story.

1 · Add a Character

Find Character in the left sidebar under Add Nodes and click +. Select the node and fill in the properties panel — be concrete; wardrobe and posture beat adjectives:

The Ray Character node selected on the canvas beside the Properties panel showing his name and full description
The node on the canvas, the details in Properties — type it once, reuse it everywhere.

2 · Add a Style

3 · Add a Time Period

Set Era to 1990s, Region to American South, and add notes: 1990 Atlanta underground music scene. The intersection of punk, hardcore, and early grunge. DIY venue culture. The era gets woven into every prompt this node touches.

4 · Add a Negative

What to keep out of the frame:

5 · Add an Output, then connect

Add an Output node — it uses the model and aspect ratio from your Settings defaults. Optionally, add a Parameters node to override them for this Output (here, 2:3 for a full-body portrait). Overrides are per-Output, so one canvas can generate with different models. Wire it up:

Character ──┐
Style ──────┼──→ Output ←── Negative
            │      ↑
Time Period ┴──────┤ (config)
Parameters ────────┘ (config)
The complete Ray portrait graph: Time Period, Character, Style, Negative, and Parameters nodes wired into an Output showing the assembled prompt and a generated full-body portrait
The whole graph — five nodes feeding one Output. The assembled prompt is visible on the Output before you spend anything.

6 · Generate

Check the assembled prompt on the Output node — it shows exactly what the graph built, era first, before you generate. Pick your model (or Mock, free, to test the wiring), then click Generate.

The Output node up close: assembled prompt on top, the generated full-body portrait of Ray below, Generate button at the bottom
The payoff — Ray, exactly as described, ready to become a Reference for every scene he's in.

Key concepts

Node types deconstruct the story. The type system is how you take a narrative apart into reusable pieces: who (Characters), where (Settings), what things (Props), how it looks (Style, Camera, Time Period) — those are built once and reused everywhere — while what happens (Actions, Shots) changes scene to scene. Organize a story this way and every new scene is mostly wiring, not rewriting.

Pull-based assembly. Only nodes connected to an Output contribute to its prompt. Everything else on the canvas is just an available asset — build a library and wire pieces in per shot.

A real FlowBoard scene graph: reference images, characters, and settings wired through actions into two Output nodes
A real scene graph — everything wired to an Output contributes; the rest waits on the canvas.

Every node, explained. This manual's node glossary covers all 24 node types — what each does, when to reach for it, and what it pairs with. Keep it open in a second tab while you learn.

Highlight script text into nodes. With a screenplay open in the Script Panel, select any passage, then click a node type in the sidebar: FlowBoard drops that node on the canvas with your selection pasted in — a selection like **Ray** — club manager of The Point even splits into name and description. The script text stays linked to the node, so deconstructing a story is mostly highlighting.

Grouping. Select nodes and press ⌘G to group them (⌘⇧G to ungroup). Groups move together and get a labeled bounding box. Click the button just inside a group's bottom-left corner to isolate it: everything else fades, and — the real point — the group unlocks, so you can rearrange its members individually. Press Escape (or click ◎ again) to lock it back into moving as one.

A FlowBoard canvas with three labeled groups — 'Edit Hero', 'Edit Villian', and 'Animate' — each with a colored dashed bounding box around its node cluster and a small ◎ isolate button at the bottom-left corner
Three groups, three jobs — each cluster moves as one, labeled and color-coded, with its ◎ isolate toggle at the bottom-left.

Compose a page

Once a few shots are generated:

  1. Add a Page node.
  2. Wire each Output into a panel slot.
  3. Pick a preset layout (2-up, 4-up, manga styles) or a dynamic grid; set gutter and export size.
  4. Click Export Page for a finished PNG.
Two reference images flowing through Transform crop nodes into a Page node showing a 2-up vertical preview with the Export Page button
Two frames, two crops, one page — References through Transforms into a Page node, ready to export.
An exported comic page: wide establishing shot of a Victorian house at night over two panels — a grandmother and boy at the door, and a red Ferrari at the curb
What Export Page produces — a composed 3-up page, straight from the app.

See Page, Transform (for cropping images inside panels), and the Skills chapter for the automated version of this whole flow.

Tips

Keyboard shortcuts

Shortcut Action
⌘G Group selected nodes
⌘⇧G Ungroup
⌘C / ⌘V Copy / paste nodes
⌘S Save project
Escape Exit isolation mode
Delete Delete selected nodes

Troubleshooting

"Out of credits." Generation pauses until you top up — Buy Credits on your dashboard — or add your own API key for that provider in Settings. Your graph and images are untouched.

Generate button shows a higher cost than expected. Cost scales with model choice and, for video, duration. Check the Parameters node's model and the Video node's duration.

Wire connects to the wrong slot. After changing a Page node's panel count, deselect and reselect the node to refresh its handles.

Image not appearing in a Page panel. The Output feeding that panel needs a generated image first — check its preview.

Next