For everyone: your first pictures

This guide assumes nothing. If a word is unfamiliar, it gets explained the first time it appears.

What this app actually is

You write a list of pictures you want, one per line, in something that looks like a spreadsheet. You press a button. It makes them all and puts them in a folder with tidy names.

That is the whole idea. Everything else is a convenience on top of it.

If you would rather just describe one picture in a sentence, there is a Chat for that too — and it can write a whole list of rows for you.

Words you will meet

API key — a long password that lets this app use somebody else’s picture machine on your behalf. You get one from the company that runs the machine. Treat it like a password: anyone who has it can spend your allowance.

Engine — the company or program that actually draws the picture. OVHcloud, Cloudflare, Google and Pollinations are engines. So is your own computer.

Model — a specific painter inside an engine. Different models have different strengths and prices.

Manifest — the spreadsheet. One row per picture.

Prompt — the sentence describing what you want drawn.

Seed — a number that makes randomness repeatable. Same prompt plus same seed gives you the same picture again, on engines that accept a seed.

Make a picture with nothing set up

  1. In the toolbar at the top, click the engine button.
  2. Choose OVHcloud · SDXL. It needs no key and no account.
  3. Click Add a picture, type a description in prompt, and press Run queue.

It makes two pictures a minute and they are always square. That is enough to learn the app and to make real pictures.

Just want to see where the buttons are? Choose Simulated Forge instead. It draws simple scenes on your own computer, instantly — not AI pictures, but everything else behaves exactly as it will with a real engine.

A faster free engine: Cloudflare

Cloudflare Workers AI gives about 690 images a day, no credit card, and is quicker than OVHcloud.

Step 1 — make a Cloudflare account

Go to dash.cloudflare.com/sign-up and sign up. Email and password. You do not need a domain name and you do not need to enter a card.

Step 2 — find your Account ID

Once you are signed in, look at the address bar. The URL looks like:

https://dash.cloudflare.com/abc123def456.../home

That long string of letters and numbers after .com/ is your Account ID. Copy it.

Step 3 — make an API token

  1. Click your profile icon (top right) → Profile
  2. Left menu → API Tokens
  3. Create Token
  4. Scroll to the bottom and choose Create Custom Token → Get started
  5. Give it a name, e.g. image forge
  6. Under Permissions, set the three dropdowns to: Account · Workers AI · Read
  7. Continue to summary → Create Token

The token appears once. Copy it now. If you lose it, delete that token and make another; nothing bad happens.

Step 4 — put both into the app

Settings → Image engines → Cloudflare. Paste the Account ID into the first box and the token into the second, then press Check the account and token. If it does not work, the message tells you which of the two is wrong.

Then pick Cloudflare · FLUX in the toolbar’s engine button.

Your first real batch

  1. Click Add a picture.

  2. In filename, type something like image_bakery.png

    The app checks your filenames as you type — lowercase, underscores instead of spaces, starting with what the row makes (image_). If one breaks a rule it says which, and auto-fix repairs it. You can switch most rules off in Settings → Filenames if your project names things differently.

  3. In prompt, describe the picture: “a warm village bakery at dawn, bread in the window, soft morning light”
  4. Press Run queue.

The row shows a thumbnail when it is done. Click the row, then the picture, to see it full size.

The file may end in .jpg rather than .png. Some engines send JPEG, and the app names the file for what it really is.

Keeping the pictures

By default they live inside the app. To get them onto your disk, pick one:

Link folder (best) — choose a folder, and every picture from then on is written there as it finishes, sorted into images/, vectors/, lottie/, sheets/ and gifs/.

Free cloud backup: point it at your OneDrive or Google Drive folder. Everything syncs itself with no extra work.

Download — a ZIP with all the pictures, the folders and the CSV inside.

Save — one picture at a time, from the row.

Keeping your keys

Settings shows Saved and checked at… and how many keys are stored. Press Back up to a file once your keys are in, and keep that file somewhere private — it holds them in plain text. Restore reads it back.

About money

You cannot spend money by accident. Free engines never ask, because there is nothing to ask about. A paid engine always stops and shows you a dialog first, like:

2 pictures on Nano Banana 2 Lite — about $0.067 ($0.034 each). The other 7 are free.

You press Spend it — go ahead, or pick one of the free engines listed underneath. Nothing is spent until you press the first one.

When it goes wrong

“Every key is resting” — you used up the allowance. The rows park themselves and try again automatically when it resets. Or switch to a different engine and press Run queue again.

“OVHcloud is still busy” — it allows two pictures a minute. The app waits its turn automatically; this only appears if it stayed busy anyway. Try again shortly.

Google says “prepayment credits are depleted” — the key is fine; the account behind it has no picture balance. This is very common and counter-intuitive. The full explanation is here.

A picture came out wrong — open the row, write what was wrong in the Note field in plain words — “too dark, and she should be facing left” — and press Generate again. Your note is folded into the prompt.

Text in the picture is gibberish — most image models cannot spell. Use a Nano Banana model for anything with real words in it, or add the text yourself afterwards with the Letterer, which uses actual fonts.

Where to go next