API

{
  "1": {
    "inputs": {
      "model": "skyreels_hunyuan_i2v_bf16.safetensors",
      "base_precision": "bf16",
      "quantization": "disabled",
      "load_device": "offload_device",
      "attention_mode": "sageattn_varlen",
      "auto_cpu_offload": false,
      "upcast_rope": true,
      "compile_args": [
        "57",
        0
      ],
      "block_swap_args": [
        "59",
        0
      ],
      "lora": [
        "80",
        0
      ]
    },
    "class_type": "HyVideoModelLoader",
    "_meta": {
      "title": "HunyuanVideo Model Loader"
    }
  },
  "3": {
    "inputs": {
      "width": [
        "45",
        1
      ],
      "height": [
        "45",
        2
      ],
      "num_frames": 97,
      "steps": 30,
      "embedded_guidance_scale": 1,
      "flow_shift": 9,
      "seed": 15,
      "force_offload": 1,
      "denoise_strength": 1,
      "scheduler": "SDE-DPMSolverMultistepScheduler",
      "model": [
        "1",
        0
      ],
      "hyvid_embeds": [
        "30",
        0
      ],
      "stg_args": [
        "43",
        0
      ],
      "teacache_args": [
        "64",
        0
      ]
    },
    "class_type": "HyVideoSampler",
    "_meta": {
      "title": "HunyuanVideo Sampler"
    }
  },
  "5": {
    "inputs": {
      "enable_vae_tiling": true,
      "temporal_tiling_sample_size": 64,
      "spatial_tile_sample_min_size": 192,
      "auto_tile_size": false,
      "vae": [
        "7",
        0
      ],
      "samples": [
        "3",
        0
      ]
    },
    "class_type": "HyVideoDecode",
    "_meta": {
      "title": "HunyuanVideo Decode"
    }
  },
  "7": {
    "inputs": {
      "model_name": "hunyuan_video_vae_fp32_kijai.safetensors",
      "precision": "fp32",
      "compile_args": [
        "57",
        0
      ]
    },
    "class_type": "HyVideoVAELoader",
    "_meta": {
      "title": "HunyuanVideo VAE Loader"
    }
  },
  "16": {
    "inputs": {
      "llm_model": "Kijai/llava-llama-3-8b-text-encoder-tokenizer",
      "clip_model": "openai/clip-vit-large-patch14",
      "precision": "bf16",
      "apply_final_norm": false,
      "hidden_state_skip_layer": 3,
      "quantization": "disabled",
      "load_device": "offload_device"
    },
    "class_type": "DownloadAndLoadHyVideoTextEncoder",
    "_meta": {
      "title": "(Down)Load HunyuanVideo TextEncoder"
    }
  },
  "30": {
    "inputs": {
      "prompt": "Cinematic scene shows a woman getting up and walking away.The background remains consistent throughout the scene. ",
      "force_offload": "bad quality video",
      "prompt_template": "video",
      "text_encoders": [
        "16",
        0
      ],
      "custom_prompt_template": [
        "88",
        0
      ],
      "hyvid_cfg": [
        "58",
        0
      ]
    },
    "class_type": "HyVideoTextEncode",
    "_meta": {
      "title": "HunyuanVideo TextEncode"
    }
  },
  "34": {
    "inputs": {
      "frame_rate": 24,
      "loop_count": 0,
      "filename_prefix": "HunyuanVideo_skyreel_I2V",
      "format": "video/h264-mp4",
      "pix_fmt": "yuv420p",
      "crf": 19,
      "save_metadata": true,
      "trim_to_audio": false,
      "pingpong": false,
      "save_output": true,
      "images": [
        "52",
        0
      ]
    },
    "class_type": "VHS_VideoCombine",
    "_meta": {
      "title": "Video Combine 🎥🅥🅗🅢"
    }
  },
  "43": {
    "inputs": {
      "enable_vae_tiling": false,
      "temporal_tiling_sample_size": 64,
      "spatial_tile_sample_min_size": 256,
      "auto_tile_size": true,
      "noise_aug_strength": 0.04,
      "latent_strength": 1,
      "vae": [
        "7",
        0
      ],
      "image": [
        "45",
        0
      ]
    },
    "class_type": "HyVideoEncode",
    "_meta": {
      "title": "HunyuanVideo Encode"
    }
  },
  "45": {
    "inputs": {
      "width": 544,
      "height": 960,
      "upscale_method": "lanczos",
      "keep_proportion": false,
      "divisible_by": 2,
      "crop": "center",
      "image": [
        "79",
        0
      ]
    },
    "class_type": "ImageResizeKJ",
    "_meta": {
      "title": "Resize Image"
    }
  },
  "52": {
    "inputs": {
      "inputcount": 2,
      "direction": "right",
      "match_image_size": false,
      "Update inputs": null,
      "image_1": [
        "45",
        0
      ],
      "image_2": [
        "60",
        0
      ]
    },
    "class_type": "ImageConcatMulti",
    "_meta": {
      "title": "Image Concatenate Multi"
    }
  },
  "57": {
    "inputs": {
      "backend": "inductor",
      "fullgraph": false,
      "mode": "default",
      "dynamic": false,
      "dynamo_cache_size_limit": 64,
      "compile_single_blocks": true,
      "compile_double_blocks": true,
      "compile_txt_in": false,
      "compile_vector_in": false,
      "compile_final_layer": false
    },
    "class_type": "HyVideoTorchCompileSettings",
    "_meta": {
      "title": "HunyuanVideo Torch Compile Settings"
    }
  },
  "58": {
    "inputs": {
      "negative_prompt": "camera movement, jump cut, scene cut, transition, fading, morphing",
      "cfg": 6,
      "start_percent": 0,
      "end_percent": 0.5
    },
    "class_type": "HyVideoCFG",
    "_meta": {
      "title": "HunyuanVideo CFG"
    }
  },
  "59": {
    "inputs": {
      "double_blocks_to_swap": 20,
      "single_blocks_to_swap": 10,
      "offload_txt_in": false,
      "offload_img_in": false
    },
    "class_type": "HyVideoBlockSwap",
    "_meta": {
      "title": "HunyuanVideo BlockSwap"
    }
  },
  "60": {
    "inputs": {
      "method": "mkl",
      "strength": 1,
      "image_ref": [
        "45",
        0
      ],
      "image_target": [
        "5",
        0
      ]
    },
    "class_type": "ColorMatch",
    "_meta": {
      "title": "Color Match"
    }
  },
  "64": {
    "inputs": {
      "weight": 4,
      "single_blocks": true,
      "double_blocks": true,
      "start_percent": 0,
      "end_percent": 1
    },
    "class_type": "HyVideoEnhanceAVideo",
    "_meta": {
      "title": "HunyuanVideo Enhance A Video"
    }
  },
  "78": {
    "inputs": {
      "frame_rate": 24,
      "loop_count": 0,
      "filename_prefix": "HunyuanVideo_skyreel_I2V",
      "format": "video/h264-mp4",
      "pix_fmt": "yuv420p",
      "crf": 19,
      "save_metadata": true,
      "trim_to_audio": false,
      "pingpong": false,
      "save_output": true,
      "images": [
        "60",
        0
      ]
    },
    "class_type": "VHS_VideoCombine",
    "_meta": {
      "title": "Video Combine 🎥🅥🅗🅢"
    }
  },
  "79": {
    "inputs": {
      "image": "image__01418_.png",
      "upload": "image"
    },
    "class_type": "LoadImage",
    "_meta": {
      "title": "Load Image"
    }
  },
  "80": {
    "inputs": {
      "lora": "hunyuan_video/skyreels-i2v-smooth-lora-test-00000350.safetensors",
      "strength": 0.6
    },
    "class_type": "HyVideoLoraSelect",
    "_meta": {
      "title": "HunyuanVideo Lora Select"
    }
  },
  "88": {
    "inputs": {
      "custom_prompt_template": "<|start_header_id|>system<|end_header_id|>\n\nYou are a professional Content Analyst. As a specialist Scene Annotator, you tag and describe scenes in detail, paying special attention to the temporal coherence and sequence of events in the scene.\n\n# INSTRUCTIONS\n\nYou will receive two inputs:\n\n1. The first frame of a video;\n2. A short description of the scene.\n\nYour job is to write a comprehensive description of the scene based on the inputs.\n\n# IMPORTANT INFORMATION \n\nThe scene is only 4s (four seconds long).\nThe scene has 98 frames in total at 24 frames per second (24 fps).\n\n# GUIDELINES\n\n- Write a detailed description based on the inputs.\n- Use your expertise to adapt the scene description so that it is consistent and coherent.\n- Ensure the actions and events described can fit reasonably in a 4 second clip.\n- Be concise and avoid abstract qualifiers, focus on concrete aspects of the scene, in particular the main subject, the motion, actions and events.\n- Use the input image, which is the fisrt frame of the video, to infer the visual qualities, mood and tone of the overall scene.\n- Always consider the appropriate sequence of events for the scene so it fits the four seconds legth of the clip.\n\n# DELIVERABLES\n\nYou will deliver a concise and detailed description of the scene that is consistent with the inputs you receive, and temporally coherent given the length of the scene. You should output something like this:\n\nDETAILED DESCRIPTION OF THE VIDEO SCENE IN APPROPRIATE TEMPORAL SEQUENCE. OVERALL MOOD OF THE SCENE BASED ON THE FIRST FRAME. 3 TO 5 TAGS THAT REPRESENT THE SCENE GENRE/STYLE/CATEGORY.\n\n\nWrite the scene description: \n\n<|eot_id|><|start_header_id|>user<|end_header_id|>\n\n{}<|eot_id|>",
      "crop_start": 95
    },
    "class_type": "HyVideoCustomPromptTemplate",
    "_meta": {
      "title": "HunyuanVideo Custom Prompt Template"
    }
  }
}