LoRAs and custom workflows in LTX Desktop

LTX Desktop is built on the open-source LTX-2.3 model, which means you can extend it with LoRAs and IC-LoRAs from the LTX community. This is aimed at users comfortable with the LTX open-source ecosystem.

What are LoRAs?

LoRAs (Low-Rank Adaptations) are lightweight model add-ons — typically 1–128MB — that steer LTX-2 generation toward a specific style, effect, or motion type without retraining the full model. You load a LoRA on top of the base model and it shifts output in the direction it was trained on.

What LoRAs can do:

• Lock generation to a specific visual aesthetic or cinematic look
• Improve consistency for a specific character or object
• Fine-tune how motion is interpreted (e.g., camera dolly moves)
• Add structural control via depth maps, pose skeletons, or edge detection

LTX provides a library of official LoRAs for common camera movements (dolly in/out, jib up/down, static), and the community publishes many more. For full details on how LoRAs work, adjusting strength, and training your own, see the LoRA documentation.

What are IC-LoRAs?

IC-LoRAs (In-Context LoRAs) are a specialized type of adapter that conditions video generation on a reference input — an existing video clip, a depth map, a pose skeleton, or another control signal. Where a standard LoRA modifies style globally, an IC-LoRA performs a targeted, reference-driven operation: structural control, VFX, video restoration, or creative transformation.

Some examples of what IC-LoRAs can do:

• Follow a depth map or pose skeleton to control scene structure
• Add effects like water simulation or day-to-night transformation
• Deblur, colorize, or otherwise restore reference footage
• Control motion using sparse keypoint trajectories

IC-LoRAs require a reference input alongside your text prompt, and most use ComfyUI for the full workflow. For a complete list of available IC-LoRA adapters and how to use them, see the IC-LoRA documentation.

Adding a LoRA to LTX Desktop

1. Download a compatible LTX LoRA (see community resources below).
2. Place the LoRA file in your LTX Desktop models folder:
   • Windows: %LOCALAPPDATA%\LTXDesktop\models\
   • macOS: ~/Library/Application Support/LTXDesktop/models/
   • Linux: ~/.local/share/LTXDesktop/models/
3. Restart LTX Desktop.
4. The LoRA will appear as a selectable option in the model/style selector in Gen Space.

Community resources

The LTX community maintains a growing library of LoRAs, IC-LoRAs, and workflow templates:

• LTX Community — browse and download community LoRAs and workflow submissions
• LoRA documentation — adjusting strength, training custom LoRAs, available official LoRAs
• IC-LoRA documentation — full adapter list, ComfyUI setup, control signal preparation

For questions, workflow sharing, and troubleshooting, the LTX Discord is the most active resource: Join the LTX Discord.