skills/ai-video-production-master/workflows/comfyui_i2v_optimized.json

{
  "name": "Wan 2.1 I2V Optimized for Apple Silicon",
  "description": "Optimized Image-to-Video workflow for M4 Max with GGUF quantization",
  "author": "ai-video-production-master skill",
  "version": "1.0.0",
  "nodes": {
    "1": {
      "class_type": "LoadImage",
      "inputs": {
        "image": "input.png"
      },
      "_meta": {
        "title": "Load Keyframe"
      }
    },
    "2": {
      "class_type": "ImageScale",
      "inputs": {
        "image": ["1", 0],
        "width": 832,
        "height": 480,
        "upscale_method": "lanczos",
        "crop": "center"
      },
      "_meta": {
        "title": "Scale to 480p (optimal for I2V)"
      }
    },
    "3": {
      "class_type": "UnetLoaderGGUF",
      "inputs": {
        "unet_name": "wan2.1-i2v-14b-480p-Q5_K_M.gguf"
      },
      "_meta": {
        "title": "Load Wan 2.1 I2V GGUF (Q5_K_M)"
      }
    },
    "4": {
      "class_type": "CLIPLoader",
      "inputs": {
        "clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
        "type": "wan"
      },
      "_meta": {
        "title": "Load UMT5 Text Encoder"
      }
    },
    "5": {
      "class_type": "VAELoader",
      "inputs": {
        "vae_name": "wan_2.1_vae.safetensors"
      },
      "_meta": {
        "title": "Load Wan VAE"
      }
    },
    "6": {
      "class_type": "CLIPTextEncode",
      "inputs": {
        "text": "subtle organic motion, gentle breathing, cinematic quality",
        "clip": ["4", 0]
      },
      "_meta": {
        "title": "Positive Prompt (Motion Description)"
      }
    },
    "7": {
      "class_type": "CLIPTextEncode",
      "inputs": {
        "text": "blurry, distorted, low quality, watermark, text, static, frozen",
        "clip": ["4", 0]
      },
      "_meta": {
        "title": "Negative Prompt"
      }
    },
    "8": {
      "class_type": "WanImageToVideo",
      "inputs": {
        "positive": ["6", 0],
        "negative": ["7", 0],
        "vae": ["5", 0],
        "width": 832,
        "height": 480,
        "length": 33,
        "batch_size": 1,
        "start_image": ["2", 0]
      },
      "_meta": {
        "title": "Wan I2V Conditioning (33 frames = ~2 sec at 16fps)"
      }
    },
    "9": {
      "class_type": "ModelSamplingSD3",
      "inputs": {
        "model": ["3", 0],
        "shift": 8.0
      },
      "_meta": {
        "title": "Apply Flow Matching (shift=8 for I2V)"
      }
    },
    "10": {
      "class_type": "KSampler",
      "inputs": {
        "model": ["9", 0],
        "positive": ["8", 0],
        "negative": ["8", 1],
        "latent_image": ["8", 2],
        "seed": 42,
        "steps": 6,
        "cfg": 5.0,
        "sampler_name": "uni_pc",
        "scheduler": "normal",
        "denoise": 1.0
      },
      "_meta": {
        "title": "Sample (6 steps optimal for speed/quality)"
      }
    },
    "11": {
      "class_type": "VAEDecode",
      "inputs": {
        "samples": ["10", 0],
        "vae": ["5", 0]
      },
      "_meta": {
        "title": "Decode Latents to Frames"
      }
    },
    "12": {
      "class_type": "VHS_VideoCombine",
      "inputs": {
        "frame_rate": 16,
        "loop_count": 0,
        "filename_prefix": "wan_i2v",
        "format": "video/h264-mp4",
        "pingpong": false,
        "save_output": true,
        "images": ["11", 0]
      },
      "_meta": {
        "title": "Export MP4 (16fps)"
      }
    }
  },
  "parameters": {
    "motion_prompt": {
      "description": "Describe the desired motion",
      "default": "subtle organic motion, gentle breathing",
      "node": "6",
      "input": "text"
    },
    "negative_prompt": {
      "description": "What to avoid in the output",
      "default": "blurry, distorted, low quality, watermark, text, static, frozen",
      "node": "7",
      "input": "text"
    },
    "steps": {
      "description": "Sampling steps (more = better quality, slower)",
      "default": 6,
      "range": [4, 12],
      "node": "10",
      "input": "steps"
    },
    "cfg": {
      "description": "Classifier-free guidance scale",
      "default": 5.0,
      "range": [3.0, 8.0],
      "node": "10",
      "input": "cfg"
    },
    "frames": {
      "description": "Number of output frames (16fps)",
      "default": 33,
      "range": [17, 81],
      "node": "8",
      "input": "length"
    },
    "shift": {
      "description": "Flow matching shift (8 for I2V, 5 for T2V)",
      "default": 8.0,
      "range": [5.0, 12.0],
      "node": "9",
      "input": "shift"
    }
  },
  "performance_notes": {
    "apple_silicon": {
      "m4_max_128gb": {
        "steps_6": "~90 minutes per clip",
        "steps_4": "~60 minutes per clip",
        "memory_usage": "~12-15GB unified memory",
        "recommendation": "Use for iteration/testing, cloud for batch"
      }
    },
    "cloud_gpu": {
      "h100_80gb": {
        "steps_6": "~2 minutes per clip",
        "steps_12": "~4 minutes per clip",
        "cost": "$1.87-2.99/hr on Vast.ai/RunPod"
      },
      "a100_80gb": {
        "steps_6": "~3 minutes per clip",
        "steps_12": "~6 minutes per clip",
        "cost": "$1.50-1.74/hr"
      }
    }
  },
  "required_models": {
    "unet": {
      "name": "wan2.1-i2v-14b-480p-Q5_K_M.gguf",
      "size": "12.7 GB",
      "source": "https://huggingface.co/city96/Wan2.1-I2V-14B-480P-GGUF",
      "path": "models/diffusion_models/"
    },
    "clip": {
      "name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors",
      "size": "~5 GB",
      "source": "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged",
      "path": "models/text_encoders/"
    },
    "vae": {
      "name": "wan_2.1_vae.safetensors",
      "size": "~200 MB",
      "source": "https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged",
      "path": "models/vae/"
    }
  },
  "required_nodes": [
    "ComfyUI-GGUF",
    "ComfyUI-VideoHelperSuite"
  ]
}