<h1 id="nsfw-multi-stage-titjob-and-cumshot-scene-generator">NSFW Multi-Stage Titjob &amp; Cumshot Scene Generator</h1>This ComfyUI workflow is designed to generate a long, multi-stage, explicit video sequence from a single starting image of a woman. Its unique, chained structure creates a narrative that progresses through several distinct sexual acts, with each scene building upon the last.<h2 id="how-it-works">How It Works</h2>The workflow operates on the principle of sequential frame generation. Instead of creating one long video at once, it divides the process into several key stages, each with its own dedicated prompts and LoRA models to guide the action:<ol><li>Scene 1 (Undressing): The process begins with a starting image of a woman. The first set of prompts and LoRAs guides the AI to generate a video where she undresses, revealing her breasts.</li><li>Frame Handoff: The last frame from the generated scene is automatically selected and becomes the starting image for the next stage, ensuring a seamless transition.</li><li>Scene 2 &amp; 3 (The Act): The process repeats with new sets of prompts. Based on the final frame of the previous scene, the character is now guided to kneel before an erect penis and perform a titjob.</li><li>Scene 4 (Climax): The final stage uses new, specialized LoRA models and prompts designed specifically to create the finale: a penis ejaculating a large amount of semen onto her face and breasts.</li><li>Fluidity Improvement (Interpolation): After each main scene is generated, the RIFE (VFI) node is used for frame interpolation. It smooths the motion and increases the frames per second (FPS), making the final video much more fluid.</li><li>Video Combining: At the very end, all the generated and improved scenes are combined into a single MP4 video file using a node from the Video Helper Suite package.</li></ol>This approach allows for the creation of a complete, explicit narrative video that changes dynamically while maintaining visual consistency.<h2 id="a-note-on-performance">A Note on Performance</h2>Please be aware that this is a very resource-intensive workflow. The entire generation process takes approximately one hour on a setup with a GeForce 4060 Ti (8GB VRAM) and 32GB of system RAM.<h2 id="required-models-and-custom-nodes">Required Models and Custom Nodes</h2>For this workflow to function correctly, you need to download and place the following files in the appropriate ComfyUI folders and install the custom nodes.<h3 id="main-models-(unet-gguf)">Main Models (Unet GGUF)</h3><ul><li><code>Wan2.2-I2V-A14B-HighNoise-Q5_K_S.gguf</code></li><li><code>Wan2.2-I2V-A14B-LowNoise-Q5_K_S.gguf</code></li></ul><h3 id="vae">VAE</h3><ul><li><code>Wan2.1_VAE.safetensors</code></li></ul><h3 id="lora-models">LoRA Models</h3><ul><li><code>lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank32_bf16.safetensors</code></li><li><code>NSFW-22-H-e8.safetensors</code></li><li><code>NSFW-22-L-e8.safetensors</code></li><li><code>I2V - tit_fuck.safetensors</code></li><li><code>Bouncing Breasts - XL wan 480p .safetensors</code></li><li><code>Titjob - wan480p v1.0.safetensors</code></li><li><code>wan2.2_highnoise_cumshot_v.1.0.safetensors</code></li><li><code>wan2.2_lownoise_cumshot_v1.0.safetensors</code></li></ul><h3 id="custom-nodes">Custom Nodes</h3>Make sure you have the following packages installed via the ComfyUI Manager:<ul><li>ComfyUI-GGUF-Loader (for handling <code>.gguf</code> models)</li><li>Video Helper Suite (VHS) (for combining frames into a video)</li><li>comfyui-kjnodes (for the <code>LoadAndResizeImage</code> node)</li><li>ComfyUI-Image-Selector (for selecting frames)</li><li>ComfyUI_HavocsCall_Custom_Nodes (for the <code>Integer to String</code> node)</li><li>ComfyUI-Easy-Use (used for the node titled "length")</li></ul><h2 id="how-to-use">How to Use</h2><ol><li>Install Everything: Ensure all the above models and custom nodes are correctly installed.</li><li>Load the Workflow: Drag and drop the <code>.json</code> file onto the ComfyUI window.</li><li>Select a Starting Image: Find the node named <code>LoadAndResizeImage</code> and choose the image of the woman you want the video to start with.</li><li>Adjust the Prompts: Each scene (marked as groups "01", "02", "03", etc.) has its own set of <code>CLIPTextEncode</code> nodes (Positive and Negative). The prompts are already set up for the sequence described above, but you can tweak them if needed.</li><li>Video Length: Find the node titled "length". Set the total length of the generated video in frames here. This value is dynamically passed to each scene.</li><li>Check LoRA Models: Make sure all <code>LoraLoaderModelOnly</code> nodes have the correct LoRA files loaded, especially the new cumshot models for the final scene.</li><li>Start Generating: Click "Queue Prompt". The process can be long and resource-intensive as it generates many scenes sequentially.</li></ol>Done! After the process is complete, the final video will be saved in your <code>output</code> folder.

Files with SHA256: a4a713de8f555952299e477dec0bd0996406ef48341e283e6ca45ce7a3b43dc8