Instructions to use Woovee/ltx25-studio-runpod with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LTX-2
How to use Woovee/ltx25-studio-runpod with LTX-2:
# Install the LTX-2 pipelines git clone https://github.com/Lightricks/LTX-2.git cd LTX-2 uv sync --extra natten
# Download weights from this repo # Substitute filenames from this repo's "Files and versions" if they differ hf download Woovee/ltx25-studio-runpod \ diffusion_models/<distilled-transformer>.safetensors \ text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors \ vae/<video-vae>.safetensors \ vae/<audio-vae>.safetensors \ latent_upscale_models/<spatial-upsampler>.safetensors \ latent_upscale_models/<temporal-upsampler>.safetensors \ --local-dir models/ltx25-studio-runpod # DFR requires the detailing IC-LoRA (separate repo; strength is fixed at 0.5) hf download Lightricks/LTX-2.5-22b-IC-LoRA-Pixel-Spatial-Upscaler --local-dir models/LTX-2.5-22b-IC-LoRA-Pixel-Spatial-Upscaler# Distilled LTX-2.5 pipeline (fast) uv run python -m ltx_pipelines.distilled \ --transformer-path models/ltx25-studio-runpod/diffusion_models/<distilled-transformer>.safetensors \ --text-encoder-path models/ltx25-studio-runpod/text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors \ --video-vae-path models/ltx25-studio-runpod/vae/<video-vae>.safetensors \ --audio-vae-path models/ltx25-studio-runpod/vae/<audio-vae>.safetensors \ --spatial-upsampler-path models/ltx25-studio-runpod/latent_upscale_models/<spatial-upsampler>.safetensors \ --num-frames 121 \ --prompt "A beautiful sunset over the ocean" \ --output-path output.mp4 # For image-to-video, add: --image path/to/image.jpg 0 0.8# DFR pipeline (higher detail fidelity; optional temporal 2x/4x) uv run python -m ltx_pipelines.dfr_pipeline \ --transformer-path models/ltx25-studio-runpod/diffusion_models/<distilled-transformer>.safetensors \ --text-encoder-path models/ltx25-studio-runpod/text_encoders/gemma4-12b-with-proj-ltx-2.5-bf16.safetensors \ --video-vae-path models/ltx25-studio-runpod/vae/<video-vae>.safetensors \ --audio-vae-path models/ltx25-studio-runpod/vae/<audio-vae>.safetensors \ --spatial-upsampler-path models/ltx25-studio-runpod/latent_upscale_models/<spatial-upsampler>.safetensors \ --temporal-upsampler-path models/ltx25-studio-runpod/latent_upscale_models/<temporal-upsampler>.safetensors \ --detailing-lora models/LTX-2.5-22b-IC-LoRA-Pixel-Spatial-Upscaler/ltx-2.5-22b-ic-lora-pixel-spatial-upscaler-x2-1.0.safetensors \ --spatial-upscalings 1 \ --temporal-upscalings 1 \ --height 1088 \ --width 1920 \ --num-frames 121 \ --prompt "A beautiful sunset over the ocean" \ --output-path output.mp4 # For 4K: --spatial-upscalings 2 --width 3840 --height 2176 # For image-to-video, add: --image path/to/image.jpg 0 0.8 - Notebooks
- Google Colab
- Kaggle
LTX-2.5 Studio package
Selected, unmodified original Lightricks LTX-2.5 assets for a private Studio on a single 48 GB GPU. This package is not affiliated with or endorsed by Lightricks.
This is not an unrestricted license. By downloading or using these weights you must comply with the complete LTX-2.x Community License Agreement, including Section 4 and Attachment A in their entirety. All upstream use restrictions and acceptable use policies apply. Redistribution does not grant any additional rights. See the original model card in NOTICE-UPSTREAM.md.
The manifest records exact source revisions, filenames, byte lengths and SHA256 values. Weights are copied without modification. Runtime integration and GPU verification status must be read from the deployment report; publishing weights is not evidence that every workflow works.
L40S test results: four video configurations passed output validation on 2026-10-18. See benchmark and deployment notes for measured timings, the tested setup and its limitations.
Candidate defaults
- Fast: distilled INT8 ConvRot, 768x512, 97 frames at 24 FPS, fixed 8 steps and CFG 1.
- Quality: two-stage 8+3 sampling with the official x2 latent upscaler.
- Hardware target: L40S 48 GB; minimum0/maximum1 workers. 24 GB is experimental and unbenchmarked.
- Text encoder is staged and offloaded before diffusion to reduce peak VRAM.
The optional adapter catalogue distinguishes usable ordinary LoRAs from task-specific IC-LoRAs that require their own conditioning workflows. A catalogue entry does not mean an adapter is enabled, mirrored, or GPU-tested.
Creation and editing workflow library
Open the capability matrix and setup guide. The package includes 18 original official ComfyUI workflows, four editing presets using the public editing-adapter package, and a repaired text-to-audio preset. Each original graph, model and custom-node source is pinned in the manifests. These are importable UI workflows, not arbitrary graphs accepted by the hosted API.
The library covers audio-conditioned video, reference ingredients, clean plates, restoration, detail refinement, control/motion tracks, inpainting, outpainting and HDR examples. Its matrix identifies remaining gated assets, known upstream issues and untested 48 GB configurations. Advanced editing remains a separate ComfyUI setup; the hosted Studio API implements text/image/first-last video and optional Cinemagraph. The four tested configurations are documented in the benchmark notes; advanced editing workflows remain unvalidated on this GPU.
- Downloads last month
- 213
Model tree for Woovee/ltx25-studio-runpod
Base model
Lightricks/LTX-2.5