Low RAM Mode

Low RAM Mode is mainly relevant to SDXL workflows.

What It Does

When Low RAM Mode is enabled, the app loads the model for inference and releases it again afterward instead of keeping everything resident in memory.

This reduces RAM pressure and makes SDXL usable on devices that would otherwise struggle to keep the whole pipeline loaded at once.

Default Behavior

For SDXL, Low RAM Mode is enabled by default.

If your device has plenty of memory, you may choose to disable it for better speed. A practical rule of thumb is that devices with 16 GB RAM or more are the main candidates for turning it off.

Tradeoff

• Low RAM Mode on: better memory behavior, slower repeated generation
• Low RAM Mode off: faster repeated generation, higher memory usage

Recommendation

If SDXL is your priority and you are unsure what to choose, keep the default.

Only disable Low RAM Mode if your device has enough RAM and you specifically want faster generation at the cost of higher memory use.