Skip to main content
Batch Mode runs one model over many generations from a single configuration, so you set the model, the quality and the prompt once and get a grid of results instead of repeating yourself. Open it at app.percify.io/home/batch. Each item is an ordinary generation, charged at that model’s normal price and refunded if it fails.
The Percify Batch Mode screen showing the model picker, quality and aspect controls, the prompt and messaging fields, strategy presets and the Generate button with its per-image credit cost

Batch Mode with the Image tab selected. The model chip shows the price per item, and the cost line under Generate shows what one image will cost at the chosen quality.

​
What can run in a batch

Batch Mode has three tabs, and each one runs the models of that kind:

​
How to run one

  1. Pick the tab, then the model. The model chip shows its price, for example GPT Image 2 · 4 cr.
  2. Set the quality and the aspect ratio, then the number of generations with the stepper.
  3. Write the Visual description: how it should look, the subject, framing and lighting.
  4. Add Messaging if the picture has to say something, such as a headline or an offer. This field is optional.
  5. Press Generate. Results stream into the preview grid as they finish, and the counter reads done and live while it runs.
The line under the button shows the price of one item at your current settings, so the cost of the run is that number times the count.

​
Strategy presets

Five presets fill in a starting prompt for a common job: Social media ads, Banner ads, Product shots, UGC hooks and Thumbnails. Press + Save to keep your own prompt scaffold as a strategy and reuse it later.

​
Start from an image instead of a prompt

Drop an image on Drop an image to extract its prompt and press Extract prompt. Percify writes a prompt describing that image, and Use as prompt puts it in the prompt box for you to edit. Extracting a prompt costs no credits. The panel beside it takes reference material that every item in the run should follow: drop screenshots, product shots or examples under References, or switch to Brand to use a saved brand, or Flow to run a saved recipe.

​
How many run at once

Concurrency sets how many generations run in parallel. It is locked while a run is going, so set it before you press Generate. Raising it finishes the set sooner and spends credits sooner; it does not change the price of any item.

​
What happens if you close the tab

The in-app run stops. Batch Mode drives its loop from the browser, so closing the tab or navigating away ends the run. Generations already submitted finish and are still charged, and you will find them in History and in your library. If you need a batch that survives you hanging up, use the API or MCP instead: those are driven by the server, resume if interrupted, and take a credit budget. See Run a batch from the API.

​
FAQ

The in-app page takes the count you set with the stepper. The API and MCP accept 1 to 50 prompts in one batch.
Each item costs the model’s normal price, which the line under Generate shows for your settings. A run of 10 images on a 4 credit model at medium quality costs what 10 single runs would cost. Nothing is charged for a run that fails.
Yes. Stopping prevents new items from being submitted. Anything already generating finishes and is still charged.
Into the preview grid as they finish, and into your library. History lists past runs.

Run a batch from the API

Server-driven batches that survive a disconnect, with a credit budget.

Compare models in the Playground

Try a model on one prompt before running it fifty times.

How to write prompts

Subject first, then style, light, camera, length and shape.

What each action costs

Credit prices per model, and what a plan includes.
Last modified on September 23, 2026