Merge pull request Lightricks#44 from Lightricks/feature/linux

Linux support

Author: alexger

.dockerignore

@@ -0,0 +1,9 @@
+node_modules
+python-embed
+dist
+dist-electron
+release
+backend/.venv
+backend/outputs
+backend/models
+.git

.github/workflows/ci.yml

@@ -59,6 +59,8 @@ jobs:
             runner: macos-latest
           - os: windows
             runner: windows-latest
+          - os: linux
+            runner: ubuntu-latest
     defaults:
       run:
         working-directory: backend

AGENTS.md

@@ -21,8 +21,8 @@ LTX Desktop is an Electron app for AI video generation using LTX models. Three-l
 | `pnpm typecheck:py` | Python pyright only |
 | `pnpm backend:test` | Run Python pytest tests |
 | `pnpm build:frontend` | Vite frontend build only |
-| `pnpm build:mac` / `pnpm build:win` | Full platform builds |
-| `pnpm setup:dev:mac` / `pnpm setup:dev:win` | One-time dev environment setup |
+| `pnpm build` | Full platform build (auto-detects platform) |
+| `pnpm setup:dev` | One-time dev environment setup (auto-detects platform) |
 
 Run a single backend test file via pnpm: `pnpm backend:test -- tests/test_ic_lora.py`
 

README.md

@@ -1,6 +1,6 @@
 # LTX Desktop
 
-LTX Desktop is an open-source desktop app for generating videos with LTX models — locally on supported Windows NVIDIA GPUs, with an API mode for unsupported hardware and macOS.
+LTX Desktop is an open-source desktop app for generating videos with LTX models — locally on supported Windows/Linux NVIDIA GPUs, with an API mode for unsupported hardware and macOS.
 
 > **Status: Beta.** Expect breaking changes.
 > Frontend architecture is under active refactor; large UI PRs may be declined for now (see [`CONTRIBUTING.md`](docs/CONTRIBUTING.md)).
@@ -32,8 +32,9 @@ LTX Desktop is an open-source desktop app for generating videos with LTX models
 | --- | --- | --- |
 | Windows + CUDA GPU with **≥32GB VRAM** | Local generation | Downloads model weights locally |
 | Windows (no CUDA, <32GB VRAM, or unknown VRAM) | API-only | **LTX API key required** |
+| Linux + CUDA GPU with **≥32GB VRAM** | Local generation | Downloads model weights locally |
+| Linux (no CUDA, <32GB VRAM, or unknown VRAM) | API-only | **LTX API key required** |
 | macOS (Apple Silicon builds) | API-only | **LTX API key required** |
-| Linux | Not officially supported | No official builds |
 
 In API-only mode, available resolutions/durations may be limited to what the API supports.
 
@@ -46,6 +47,14 @@ In API-only mode, available resolutions/durations may be limited to what the API
 - 16GB+ RAM (32GB recommended)
 - **160GB+ free disk space** (for model weights, Python environment, and outputs)
 
+### Linux (local generation)
+
+- Ubuntu 22.04+ or similar distro (x64 or arm64)
+- NVIDIA GPU with CUDA support and **≥32GB VRAM** (more is better)
+- NVIDIA driver installed (PyTorch bundles the CUDA runtime)
+- 16GB+ RAM (32GB recommended)
+- Plenty of free disk space for model weights and outputs
+
 ### macOS (API-only)
 
 - Apple Silicon (arm64)
@@ -64,6 +73,7 @@ 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).
 
@@ -84,7 +94,7 @@ The LTX API is used for:
 - API-based video generations (required on macOS and on unsupported Windows hardware) — paid
 - Retake — paid
 
-An LTX API key is required in API-only mode, but optional on Windows local mode if you enable the Local Text Encoder.
+An LTX API key is required in API-only mode, but optional on Windows/Linux local mode if you enable the Local Text Encoder.
 
 Generate a FREE API key at the [LTX Console](https://console.ltx.video/). Text encoding is free; video generation API usage is paid. [Read more](https://ltx.io/model/model-blog/ltx-2-better-control-for-real-workflows).
 
@@ -137,11 +147,7 @@ Prereqs:
 Setup:
 
 ```bash
-# macOS
-pnpm setup:dev:mac
-
-# Windows
-pnpm setup:dev:win
+pnpm setup:dev
 ```
 
 Run:

backend/pyproject.toml

@@ -27,6 +27,7 @@ dependencies = [
     "uvicorn[standard]>=0.30.0",
     "python-multipart>=0.0.9",
     "triton-windows; sys_platform == 'win32'",
+    "triton; sys_platform == 'linux'",
 ]
 
 [[tool.uv.index]]

backend/runtime_config/runtime_policy.py

@@ -8,7 +8,7 @@ def decide_force_api_generations(system: str, cuda_available: bool, vram_gb: int
     if system == "Darwin":
         return True
 
-    if system == "Windows":
+    if system in ("Windows", "Linux"):
         if not cuda_available:
             return True
         if vram_gb is None:

backend/tests/test_runtime_policy_decision.py

@@ -26,5 +26,21 @@ def test_windows_with_required_vram_allows_local_mode() -> None:
     assert decide_force_api_generations(system="Windows", cuda_available=True, vram_gb=31) is False
 
 
+def test_linux_without_cuda_forces_api() -> None:
+    assert decide_force_api_generations(system="Linux", cuda_available=False, vram_gb=24) is True
+
+
+def test_linux_with_low_vram_forces_api() -> None:
+    assert decide_force_api_generations(system="Linux", cuda_available=True, vram_gb=30) is True
+
+
+def test_linux_with_unknown_vram_forces_api() -> None:
+    assert decide_force_api_generations(system="Linux", cuda_available=True, vram_gb=None) is True
+
+
+def test_linux_with_required_vram_allows_local_mode() -> None:
+    assert decide_force_api_generations(system="Linux", cuda_available=True, vram_gb=31) is False
+
+
 def test_other_systems_fail_closed() -> None:
-    assert decide_force_api_generations(system="Linux", cuda_available=True, vram_gb=48) is True
+    assert decide_force_api_generations(system="FreeBSD", cuda_available=True, vram_gb=48) is True

backend/uv.lock

@@ -14,11 +14,7 @@ Prereqs:
 Setup:
 
 ```bash
-# macOS
-pnpm setup:dev:mac
-
-# Windows
-pnpm setup:dev:win
+pnpm setup:dev
 ```
 
 Run:

docs/CONTRIBUTING.md

@@ -10,14 +10,15 @@ This guide explains how to build a distributable installer for **LTX Desktop**.
 The installer includes:
 - **Electron app** (React frontend + Electron shell)
 - **Embedded Python** (version from [`backend/.python-version`](../backend/.python-version)) with all dependencies pre-installed:
-  - PyTorch (CUDA on Windows, MPS on macOS)
+  - PyTorch (CUDA on Windows/Linux, MPS on macOS)
   - FastAPI, Diffusers, Transformers
   - LTX-2 inference packages
   - All other required libraries
 - **Backend Python code**
 
 **NOT bundled** (downloaded at runtime):
 - Model weights (downloaded on first run; can be large) from Hugging Face
+- On **Linux** and **Windows**: the Python environment itself is downloaded on first launch (keeps installer small)
 
 The embedded Python is **fully isolated** from the target system's Python — it lives inside `{install_dir}/resources/python/` and never modifies system settings.
 
@@ -35,67 +36,43 @@ Before building, ensure you have:
 
 - **Windows**: PowerShell 5.1+ (comes with Windows 10/11)
 - **macOS**: Xcode Command Line Tools (`xcode-select --install`)
+- **Linux**: `build-essential` (or equivalent) for native extensions
 
 ## Quick Build
 
-### macOS
 ```bash
-pnpm build:mac
-```
-
-### Windows
-```powershell
-pnpm build:win
+pnpm build
 ```
 
-This will:
+This auto-detects your platform and will:
 1. Download a standalone Python distribution (version from [`backend/.python-version`](../backend/.python-version))
-2. Install all Python dependencies (~10GB on Windows with CUDA, ~2-3GB on macOS with MPS)
+2. Install all Python dependencies (~10GB on Windows/Linux with CUDA, ~2-3GB on macOS with MPS)
 3. Build the frontend
 4. Package everything with electron-builder
-5. Create a DMG (macOS) or NSIS installer (Windows) in the `release/` folder
+5. Create a DMG (macOS), AppImage + deb (Linux), or NSIS installer (Windows) in the `release/` folder
 
 ## Build Options
 
-### macOS
-
 ```bash
 # Full build
-pnpm build:mac
+pnpm build
 
 # Skip Python setup (if already prepared)
-pnpm build:mac:skip-python
+pnpm build:skip-python
 
 # Fast rebuild (unpacked, skip Python + pnpm install)
-pnpm build:fast:mac
+pnpm build:fast
 
 # Just prepare Python environment
-pnpm prepare:python:mac
+pnpm prepare:python
 ```
 
-### Windows
-
-```powershell
-# Full build
-pnpm build:win
-
-# Skip Python setup (if already prepared)
-pnpm build:win:skip-python
-
-# Just prepare Python environment
-pnpm prepare:python:win
-
-# Fast rebuild (unpacked, skip Python + pnpm install)
-pnpm build:fast:win
-
-# Clean build
-powershell -File scripts/local-build.ps1 -Clean
-```
+All commands auto-detect the current platform (macOS, Linux, or Windows).
 
 ### Build Script Options
 
-The `local-build.sh` script accepts:
-- `--platform mac|win` — Target platform (auto-detected if omitted)
+The underlying `local-build.sh` / `local-build.ps1` scripts also accept:
+- `--platform mac|linux|win` — Target platform (auto-detected if omitted)
 - `--skip-python` — Use existing `python-embed/` directory
 - `--clean` — Remove build artifacts before starting
 - `--unpack` — Build unpacked app only (faster, no installer/DMG)
@@ -108,6 +85,15 @@ release/
   └── LTX Desktop-<version>-arm64.dmg
 ```
 
+### Linux
+```
+release/
+  ├── LTX Desktop-x86_64.AppImage
+  ├── LTX Desktop-amd64.deb
+  ├── LTX Desktop-arm64.AppImage
+  └── LTX Desktop-arm64.deb
+```
+
 ### Windows
 ```
 release/
@@ -118,7 +104,7 @@ release/
 
 Place icon files in `resources/` before building:
 - `icon.ico` — Windows (multi-size ICO: 256x256, 128x128, 64x64, 48x48, 32x32, 16x16)
-- `icon.png` — macOS (1024x1024 recommended)
+- `icon.png` — macOS and Linux (1024x1024 recommended)
 
 ## Troubleshooting
 
@@ -137,6 +123,7 @@ xattr -dr com.apple.quarantine /Applications/LTX\ Desktop.app
 ### Installer is too large
 Expected installer sizes (does not include model weights):
 - **Windows**: ~10GB (PyTorch CUDA ~2.5GB + ML libraries ~5GB + Python ~200MB + Electron ~100MB)
+- **Linux**: ~10GB (similar to Windows; PyTorch CUDA variant)
 - **macOS**: ~2-3GB (PyTorch MPS is much smaller than CUDA variant)
 
 ### Runtime / first-run issues
@@ -156,10 +143,28 @@ pnpm install
 pnpm build:frontend
 
 # 4. Build DMG
-npx electron-builder --mac
+pnpm exec electron-builder --mac
+
+# Or build unpacked app (faster, for testing)
+pnpm exec electron-builder --mac --dir
+```
+
+### Linux
+```bash
+# 1. Prepare Python environment
+bash scripts/prepare-python.sh
+
+# 2. Install dependencies
+pnpm install
+
+# 3. Build frontend
+pnpm build:frontend
+
+# 4. Build AppImage + deb
+pnpm exec electron-builder --linux
 
 # Or build unpacked app (faster, for testing)
-npx electron-builder --mac --dir
+pnpm exec electron-builder --linux --dir
 ```
 
 ### Windows
@@ -174,5 +179,5 @@ pnpm install
 pnpm build:frontend
 
 # 4. Build installer
-npx electron-builder --win
+pnpm exec electron-builder --win
 ```