ltxv2-video

# LTX-V2 19B Video Workflows ## Overview LTX-V2 (LTX-2) is a 19-billion parameter DiT-based video foundation model from Lightricks. It uses a Gemma 3 12B text encoder and supports both text-to-video (T2V) and image-to-video (I2V). Key features: - **Distilled model** for fast 8-step generation - **Two-stage pipeline**: Generate at low res, then 2x spatial upscale in latent space - **Camera control LoRAs** for cinematic movements - **Audio-video generation** in a single pass (optional) ## Models ### Checkpoint (Installed) | Component | Node | Model | Notes | |-----------|------|-------|-------| | **Checkpoint** | `CheckpointLoaderSimple` | `ltx-2-19b-distilled.safetensors` | 41GB bf16, distilled variant | ### Text Encoder (Installed) | Component | Node | Model | Notes | |-----------|------|-------|-------| | **Gemma 3** | `CLIPLoader` (type=`ltxv`) | `gemma_3_12B_it_fp4_mixed.safetensors` | 9GB FP4, in text_encoders/ | **Loading note**: The checkpoint bundles the VAE internally. The Gemma 3 text encoder loads separately. Use `CLIPLoader` with `type: "ltxv"` pointing at the `text_encoders/` directory. ### LoRAs (Installed) | LoRA | File | Purpose | |------|------|---------| | **Distilled LoRA** | `ltx2/ltx-2-19b-lora-camera-control-dolly-left.safetensors` | Camera dolly left | | **Distilled LoRA (384)** | `ltx2/ltx-2-19b-distilled-lora-384.safetensors` | Apply to base model for distilled behavior | | **Camera Dolly Left** | `ltx-2-19b-lora-camera-control-dolly-left.sa...