🔖 README

🔶 All the images used to create the videos in the gallery are generated with PimpMyPony 🔶

The new version is out today, with a brand new anime style ✨

🟣 Credits

This workflows are intended to be used with the models by taek75799 as they follow the structure of Dynamic Prompts you can find under these models:

WAN 2.2 Enhanced NSFW | SVI | camera prompt adherence (Lightning Edition) I2V and T2V fp8 GGUF

Update: after some tests, the automatic prompting also works with the models by DigitalPastel

I tested it on the v2 of his Smooth Mix Wan 2.2 I2V

This version has not Lightx2v LoRas included as the previous, the one I tested is lightx2v_I2V_14B_480p_cfg_step_distill_rank128_bf16

setted at 3.00 on High and 1.5 on Low

🔖 FULL I2V WORKFLOW LEGACY

• Removed the Tensorrt Upscaler

• Disabled SageAttention and Triton acceleration [both can be reactivated by opening the first Subgraph]

• Disabled Torch Compile on Qwen3-VL

Now the WF is compatible with older PyTorch and CUDA versions (like py2.5+cu124)

🔖 I2V WORKFLOW LIGHTX2V

Just some little changes from v1.0:

• Removed the Custom Text Box

• System Prompt for Qwen3-VL is now outside the Subgraph so anyone can modify it

• Added a LoRa Stack for Lightx2v LoRaS, the LoRaS placed in that stack will be applied to all the 4 video segments

SEED EXPLANATION:

At the bottom of every Subgraph you can see 2 different seeds, the upper one is the Noise Seed of the KSampler Node that generates the images, the other is the Qwen3-VL seed, if you fix both seeds, that Subgraph will keep the video generated. So you can, for example, generate the first 5 second video, than choose to extend it to 10 second without having to generate the first one again.

Lastly, there is anothere Qwen settings you can play with, the Preset Prompt, try Cinematic or Short Story to see the difference

🔖 FULL I2V WORKFLOW v1.0

Complete workflow that includes:

• Long Video Generation [from 5 to 20 seconds]

• Auto Prompting [Qwen3-VL]

• Upscale [Tensorrt Upscale or 2xLexicaRRDBNet, dipending on your hardware]

• Frame Interpolation [30fps and 60fps for img2vid | 24fps and 50fps for MMAudio]

• MMAudio [NSFW Unlocked]

📝 INSTRUCTIONS

➡️ Install this modified node that allows you to use an uncensored QwenVL model

git clone https://github.com/huchukato/ComfyUI-QwenVL-Mod.git ComfyUI/custom_nodes/ComfyUI-QwenVL-Mod

NOTE: if you install the QwenVL node from the Manager you will be not able to use the Abliterated model.

➡️ Restart ComfyUI (it will automatically install the required dependencies)

➡️ Load the workflow and install all the other Missing Custom Nodes

The workflow is simple to use, the only things you have to remember are the text boxes in the Subgraph:

• Custom Text Box: mostly unnecessary but I prefered to add it in case there is the need to use, for example, a trigger word for some LoRaS

• Prompt Description Box [Multilanguage ITA ENG]: Just write your idea and the LLM will do the rest, formatting the prompt in the dynamic prompt format used in the Wan 2.2 models

• Final Prompt: Just shows up the final prompt

Inside the workflow you will find all the links to download the models you will need

That's all, hope you enjoy ^^

🔖 I2V AUTOPROMPT v1.0

Basic workflow for automatic prompting

➡️ Install this modified node that allows you to use an uncensored QwenVL model

git clone https://github.com/huchukato/ComfyUI-QwenVL-Mod.git ComfyUI/custom_nodes/ComfyUI-QwenVL-Mod

NOTE: if you install the QwenVL node from the Manager you will be not able to use the Abliterated model.

➡️ Restart ComfyUI (it will automatically install the required dependencies)

➡️ Load the workflow, ComfyUI will ask you to install another 2 nodes

✍🏻 AUTO PROMPTING

➡️ From the scroll down menu select this QwenVL model

➡️ Just describe what you wanna see in the video, in Italian or English, and run the workflow, it will generate the prompt with the correct syntax, you just have to copy and past it

For any trouble feel free to comment, hope you enjoy ^^