| --- |
| tags: |
| - wan |
| - wan2.2 |
| - GSsmart |
| - T2V |
| - I2V |
| - lora |
| - comfyui |
| - video |
| - uncensored |
| base_model: Wan-AI/Wan2.2-I2V-A14B |
| instance_prompt: null |
| license: apache-2.0 |
| pipeline_tag: image-text-to-video |
| --- |
| # Wan2.2 GSsmart Uncensored |
| _Left: I2V | Right: T2V_ |
| <video controls width="1280" height="720"> |
| <source src="https://huggingface.co/rzgar/Wan2.2_LightX2V_4Step_Uncensored/resolve/main/demo/ComfyUI_00006.mp4" type="video/mp4"> |
| Your browser does not support the video tag. |
| </video> |
|
|
| <Gallery /> |
|
|
| ## Model description |
|
|
| I got tired of switching back and forth between **I2V** and **T2V** models, and swapping **LightX2V** (or other LoRAs) suited to each. When you mostly generate videos for testing or “what if” runs, that repetitive setup gets old fast. |
|
|
| These LightX2V LoRAs are meant for **both I2V and T2V**, and they are **uncensored**. If you’re in the same boat, they might save you some workflow friction. |
| ___ |
| ## What to expect |
| |
| - Tuned to understand **mechanics and motion** from explicit prompts. |
| - **Female / male body accuracy** is roughly **80–85%** on **T2V**. |
| - On **I2V**, results depend heavily on the **input image**. Bringing a male or female into the scene (a man || woman appears) depends a lot on how clear and explicit your prompt is, be descriptive about clothing, body. |
| ___ |
| ## Stacking other LoRAs |
| |
| Other well-trained LoRAs usually work fine at about **0.55–0.65** strength on top of this pair. |
| |
| **Downside:** if you **lower this LoRA’s strength**, you are also weakening the **LightX2V / speed** behavior. Without another LightX2V pair to compensate, output can get **blurry** with **broken motion**. Keep this pair at full (or near-full) strength unless you add a separate speed LoRA. |
| ___ |
| ## Files |
| |
| | File | Stage | |
| |------|--------| |
| | [Wan2.2_LightX2V_high_n54vv.safetensors](https://huggingface.co/rzgar/Wan2.2_LightX2V_4Step_Uncensored/resolve/main/Wan2.2_LightX2V_high_n54vv.safetensors) | High noise | |
| | [Wan2.2_LightX2V_low_n54vv.safetensors](https://huggingface.co/rzgar/Wan2.2_LightX2V_4Step_Uncensored/resolve/main/Wan2.2_LightX2V_low_n54vv.safetensors) | Low noise | |
| |
| ### Appearance-Only Enhancers |
| The **appearance** side of the LoRA is **intentionally tuned down** for better stacking with other LoRAs. |
| Rarely needed in I2V scenes, but for T2V these appearance-only LoRAs can enhance and compensate for the tuned-down appearances. |
| |
| | File | Variant | Strength | |
| |------------------------------------------------|------------|--------------------------------------| |
| | [male_genitalia_enhancer_high.safetensors](https://huggingface.co/rzgar/Wan2.2-Bernini-R-Motion-Enhancer-n4w-i2v/resolve/main/male_genitalia_enhancer_high.safetensors) | High noise | 0.55 | |
| | [male_genitalia_enhancer_low.safetensors ](https://huggingface.co/rzgar/Wan2.2-Bernini-R-Motion-Enhancer-n4w-i2v/resolve/main/male_genitalia_enhancer_low.safetensors) | Low noise | 0.55 | |
| | [female_genitalia_enhancer_high.safetensors](https://huggingface.co/rzgar/Wan2.2-Bernini-R-Motion-Enhancer-n4w-i2v/resolve/main/female_genitalia_enhancer_high.safetensors) | High noise | 0.5 | |
| | [ female_genitalia_enhancer_low.safetensors](https://huggingface.co/rzgar/Wan2.2-Bernini-R-Motion-Enhancer-n4w-i2v/resolve/main/female_genitalia_enhancer_low.safetensors) | Low noise | 0.5 | |
| ___ |
|
|
| ## 🙏 Acknowledgements |
| [LightX2V](https://huggingface.co/lightx2v) | [Wan-AI](https://huggingface.co/Wan-AI) |
|
|
|
|
| ## Download model |
|
|
|
|
| [Download](/rzgar/Wan2.2_LightX2V_4Step_Uncensored/tree/main) them in the Files & versions tab. |
