Why can't I select local generation in LTX Desktop?
If you only see API-labelled models in the generation panel, local generation isn't available on your current setup. The cause is usually your hardware — macOS and machines without a qualifying NVIDIA GPU can't run local generation. On supported hardware, a stalled model download or outdated drivers are the most common fixable causes.
Local generation is only available on certain machines
Local generation requires Windows or Linux with a supported NVIDIA GPU. If either of the following apply, API mode is expected and there's nothing to fix:
You're on macOS. Local GPU inference is not currently supported on macOS. The app runs, but generation always routes through the LTX API. Mac support is on the roadmap.
Your machine doesn't have a qualifying GPU. An NVIDIA GPU is required — AMD and Intel GPUs are not supported for local inference. The recommended spec is 32GB+ VRAM; lower configurations may work with in-app optimizations but aren't guaranteed. See Local mode and API mode in LTX Desktop for the full hardware table.
If your machine should support local generation
On a supported Windows or Linux machine, local generation may be disabled or not yet set up. First, check whether it's simply switched off — go to Local mode and API mode in LTX Desktop for instructions on how to enable local mode in Settings.
If local mode is enabled but local models still aren't appearing, work through the steps below.
Step 1: Check the model download is complete
Go to Settings > Models and check the download status. A partial or stalled download is the most common reason local models don't appear on qualifying hardware. Let the download finish, then restart LTX Desktop.
Step 2: Update your NVIDIA drivers
Outdated drivers can prevent CUDA from being detected, causing LTX Desktop to fall back to API mode. Download the latest drivers from nvidia.com/drivers and restart after installing.
Step 3: Collect logs and escalate
If local mode still doesn't appear after completing the steps above, collect your logs and share them along with your nvidia-smi output on the GitHub issues page. See Where are LTX Desktop logs? for how to find them.