🧠 GPT Image 2

Batch image generation with your own ChatGPT account — paste prompts, pick a ratio, press Run. PLUS MAX

Overview

The GPT Image 2 page uses your ChatGPT account to generate images. It works exactly like the Flow Image page: paste a list of prompts, choose your settings, press Run — the app works through them and saves every image to your output folder.

Detail
OutputImages only — this page has no video mode
You needOne ChatGPT account (sign in from inside the app)
Aspect ratios7 ratios: 1:1, 3:2, 4:3, 16:9, 2:3, 3:4, 9:16
Reference imagesUp to 5 per row
Plan requiredPLUS or MAX
⚠️The GPT Image 2 page requires a PLUS or MAX plan. Basic-tier accounts can browse the page but not run it — the plan-required dialog pops up on Run / Run Selected / Retry Errors.

Adding a ChatGPT Account

Open Settings → ChatGPT Account and follow three steps:

  1. Press 🌐 Browser sign-in
  2. Your browser opens — sign in to ChatGPT as you normally would
  3. That's it. Switch back to the app and the account is in the list
💡Adding a second account? Sign out of chatgpt.com in your browser first. Otherwise the browser goes straight back into the old account and the app reports a duplicate.

Add several accounts. Each one has its own quota. When an account runs out, the app moves to the next and your run keeps going instead of stopping. Each account can also use its own proxy if you need one.

Basic Settings

SettingOptionsHow to choose
Aspect ratio7 ratiosMatch where you will post: 16:9 for YouTube, 9:16 for Reels/TikTok, 1:1 for Instagram
QualityLow / Medium / HighHigh for final images, Low while you are still testing prompts
Images per prompt1 – 4More images uses more quota — picking 4 costs four times as much as picking 1
ThreadsDepends on accountsThe more accounts you add, the more threads you can pick
Save modeNone / SubfolderUse subfolders so each run stays in its own folder
Output folderPathWhere images are saved

Advanced Settings

The defaults work fine — only change these when you have a specific reason.

SettingOptionsWhen to use it
Prompt modeAuto / DirectPick Direct when you have written the prompt carefully and do not want the AI rewriting it
Delay5 – 10 secondsPause between images. Leave it as is
ReasoningNone → MaxRaise it for demanding prompts: counting objects, complex layouts, text inside the image
Web searchOn / OffTurn on when the prompt mentions a real person, place or brand
Reference folderPath + toggleTurn on to let the app attach images by filename
💡Leaving reasoning on "None" is fine for most work. Raising it mostly makes images take longer rather than look better.
ℹ️Web search makes images more true-to-life when the prompt names something real, but each image takes noticeably longer. Leave it off for purely creative prompts.

Reference Images

Each row takes up to 5 reference images. Three ways to add them:

  1. 📷 Add Images — tick the rows, then pick files from your computer
  2. Drag and drop an image straight onto a row's reference cell
  3. Attach by filename — turn on the toggle next to Reference folder

How auto-attach works: name your images after the character or location, e.g. REF_CHR_DEAN.jpg. Any prompt mentioning REF_CHR_DEAN gets that image attached automatically — no picking files row by row.

Match modeWhat it means
Exact matchOnly attaches when the prompt spells out the whole filename. Writing REF_CHR_DEAN_WEDDING will not wrongly pull in REF_CHR_DEAN.jpg
Keyword matchAttaches on part of the name too — writing dean still brings in REF_CHR_DEAN.jpg
💡Switch modes in Settings → General Settings. If your library has lots of similar names (Dean, Dean Wedding, Dean Wedding Guest…), use Exact match so you do not get extra images.
ℹ️Images you picked yourself are always kept — the app only replaces the ones it attached automatically. Rows that already finished or are queued keep their images unchanged.

The 🎭 Characters button opens the shared image library used by the Flow Image page — browse the folder and rename files right there.

Work Table & Queue

The table is identical to the Flow Image page: Checkbox · # · Task · References · Prompt · Output · Progress.

Press ➕ Add to Queue to group pending rows into a task. Three things worth knowing:

  • A task remembers the settings from when you queued it — changing the panel afterwards does not affect it
  • To change it, edit that task in Queue Manager
  • Each task runs at its own thread count
ℹ️Load Session saves the whole table: prompts, reference images, settings and the images you already made. Reopen it and carry on where you left off.

Action Buttons

ButtonWhat it does
🚀 RUNRun every pending row
▶️ Run SelectedOnly run the rows you ticked
🔄 Retry ErrorsRe-run every row showing an error
🔁 RetryButton in the Progress column — re-runs that one row
📂 Open FolderButton in the Progress column — opens the folder with the image
⏸️ PAUSEPause — images already in progress still finish
⏹️ STOPStop completely
📂 Load SessionReopen a saved session
💡Retry Errors and Retry only remake what is missing. If one image out of four failed, only that one is remade — the other three are kept and cost you nothing extra.

If Something Goes Wrong

What you seeWhat to do
Could not add accountClose other open applications and try signing in again
Account already existsThe browser is still signed in to the old account — sign out of chatgpt.com and add it again
Sign-in timed outThe browser window was closed or left too long — press sign-in again
Account rate limitedThat account is out of quota. The app switches automatically — add a few more accounts for smoother runs
Content blockedThe prompt breaks ChatGPT policy. Switching accounts will not help — reword the prompt
No active accountsNo account added, or your accounts are disabled / expired. Check the ChatGPT Account tab