Skip to content

setup-and-settings.md

Latest commit

 

History

History
69 lines (45 loc) · 3.95 KB

setup-and-settings.md

File metadata and controls

69 lines (45 loc) · 3.95 KB
Raw

Setup and settings

Platform requirements

These values match the LTX Desktop page. All platforms need 100 GB+ free disk space, up to 160 GB+ with additional LoRAs and IC-LoRAs.

Windows (local generation)

  • Windows 10/11 (64-bit)
  • NVIDIA GPU with CUDA, 24 GB+ VRAM (32 GB recommended)
  • 32 GB RAM

Linux (local generation)

  • Ubuntu 22.04+ or similar distro (x64 or arm64)
  • NVIDIA GPU with CUDA and the NVIDIA driver installed (PyTorch bundles the CUDA runtime), 24 GB+ VRAM (32 GB recommended)
  • 32 GB RAM

macOS (local generation)

  • macOS 26+ (Tahoe) for local generation. The app itself needs macOS 14 or later. On macOS 14 and 15 the app shows the Coming soon to your device notice.
  • Apple Silicon (M1 or later) with 48 GB+: full local generation
  • Apple Silicon with 32 GB: limited (540p and 5 s only)
  • Apple Silicon with less than 24 GB, or an Intel Mac: Gen Space API only. Intel Macs have no MPS backend.

Lower-spec machines

The app offers local generation on lower-spec machines, but we do not recommend them. The app uses these floors (see backend/runtime_config/runtime_policy.py and docs/local-compat-matrix.md):

  • CUDA: 9 GiB VRAM on Windows with an fp8-capable GPU, 15 GiB otherwise (Linux, non-fp8 Windows; integer GiB from total_memory // 1024**3)
  • macOS: 32 GiB total RAM (machine SKU, not free RAM at launch) and macOS 26 or later

Below a floor, local generation is unavailable and the app shows the Coming soon to your device notice. An LTX API key is not required.

First run & data locations

LTX Desktop stores app data (settings, models, logs) in:

  • Windows: %LOCALAPPDATA%\LTXDesktop\
  • macOS: ~/Library/Application Support/LTXDesktop/
  • Linux: $XDG_DATA_HOME/LTXDesktop/ (default: ~/.local/share/LTXDesktop/)

Model weights are downloaded into the models/ subfolder (this can be large and may take time).

On first launch, the Setup screen guides you through the steps below. It requires internet.

  1. Sign in to Hugging Face and accept the license terms of gated models (the 2.5 weights are gated).
  2. Optionally add an LTX API key for free cloud text encoding.
  3. Choose the models to download and the models folder.
  4. Optionally select Opt out of basic analytics.

The app shows a splash screen while it installs its Python environment.

Text encoding: to generate videos you must configure text encoding:

  • LTX API key (cloud text encoding) — text encoding via the API is completely FREE and highly recommended to speed up inference and save memory. Generate a free API key at the LTX Console. Read more.
  • Local Text Encoder (extra download; enables fully-local operation on supported hardware) — if you don't wish to generate an API key, you can download the local encoder in Settings > Models and select it in Settings > General > Text encoding.

Settings

Open Settings from the bottom of the sidebar.

Tab What it controls
General Appearance (theme: system, light, or dark); auto-enhance for local and API generation; fast decode (faster video decode with slightly lower fidelity); text encoding (LTX API or local encoder); NVIDIA performance options (Torch Compile and diffusion stage cache, experimental); anonymous analytics
Models Download and manage checkpoints, LoRAs, and encoder models. See Managing models. Hidden in API-only mode
Legacy Options for Gen Space and the Video Editor: project assets folder, default base model, preference for cloud video and image generation, and seed lock
API Keys Optional LTX, fal, and Gemini keys. The app checks a key before it saves it, and you can remove a saved key
About App version, the automatic update switch, and manual update checks