For Invoke-hosted models (like SDXL, Flux, or Bria), 1 processing credit = 1 second of compute time. Compute time is calculated based on the actual time our servers spend generating images and running models using our resources.
This includes any task where our system is actively creating or processing data, such as:
- Generating images based on your input
- Running Workflows that generate images
- Running model training jobs
What does not count as compute time?
We do not charge processing credits for activities within the application that do not require substantial computational resources.
These include:
- Designing, editing, or setting up workflows.
- Uploading or managing files and assets.
- Planning or configuring settings.
- Browsing through the application or idle time within the studio space.
For API-based models (like Imagen 3 or ChatGPT 4o), credit usage is fixed per request and varies by model.
See our Support page for more information.