Unlock stunning visuals using Fluxmania V, a high-quality Text-to-Image workflow designed for ComfyUI. Utilizing a combination of FluxGuidance, DualCLIP, and T5XXL, this setup brings unmatched detail, artistic control, and storytelling capabilities to your image generation process.

---

🎯 Features

• Fluxmania V Model Integration – Powerful fine-tuned model for high-detail generations.
• DualCLIP System – Utilizes both vision and language encoders for enhanced prompt understanding.
• FluxGuidance Node – Steers generation with added semantic control.
• FP8 Model Support – Lightweight yet high-performing model files for efficient GPU usage.
• Easy GPU Cleanup – Automatically frees GPU memory after generation.
• Custom Prompt Crafting – Create emotionally rich and cinematic images with precise control.

---

💡 Use Cases

• Cinematic Portraits – Perfect for generating story-driven character visuals.
• Concept Art – Artists can iterate on mood, lighting, and style with prompt-based direction.
• AI Photography – Mimic camera settings and lenses for photorealistic compositions.
• Creative Writing & Storyboards – Visualize scenes or characters directly from prose.

---

⚙️ How It Works

1. Load Required Models

• Download fluxmania_V.safetensors into ComfyUI/models/diffusion_models.
• Add t5xxl_fp8_e4m3fn_scaled.safetensors and clipLCLIPGFullFP32_zer0intVisionCLIPL.safetensors into ComfyUI/models/clip.
• Place ae.safetensors into ComfyUI/models/vae.

2. Initialize the Workflow

• Load the provided JSON file in ComfyUI.
• Ensure all model paths are correctly mapped.

3. Input Prompts

• Positive prompt: describe the desired scene or character vividly (e.g., portrait settings, lighting, mood).
• Negative prompt: leave empty or refine for clarity.

4. Sampling & Image Generation

• Adjust the sampler and steps as needed (default: dpmpp_2m, 30 steps).
• Resolution set at 1024x1024 via EmptySD3LatentImage.

5. Save Output

• Generated image is automatically saved via SaveImage node.
• GPU memory is cleared post-generation using easy cleanGpuUsed.

---

Credits: pixaroma