From 4b2b6be3fb2ef3e7bdabbd57025571801915dd7c Mon Sep 17 00:00:00 2001
From: ethanfel <28683423+ethanfel@users.noreply.github.com>
Date: Thu, 6 Aug 2026 16:15:28 +0200
Subject: [PATCH] Add MiniMax H3 per-row inpainting workflow

---
 README.md                                     |   13 +
 .../MiniMax H3 Per-Row Video Inpaint.json     | 1694 +++++++++++++++++
 tests/test_example_workflow.py                |   73 +
 3 files changed, 1780 insertions(+)
 create mode 100644 example_workflows/MiniMax H3 Per-Row Video Inpaint.json
 create mode 100644 tests/test_example_workflow.py

diff --git a/README.md b/README.md
index bdea257..96b15fd 100644
--- a/README.md
+++ b/README.md
@@ -89,6 +89,19 @@ By default, white mask pixels are regenerated and black pixels are preserved. A
 
 The source video belongs in the main latent. Do not also pass it as an H3 `<Video>` reference merely to make masking work.
 
+## Example workflow
+
+Load [`MiniMax H3 Per-Row Video Inpaint.json`](example_workflows/MiniMax%20H3%20Per-Row%20Video%20Inpaint.json)
+in ComfyUI for a complete source-video inpainting graph. It includes source AV extraction,
+automatic `17k+5` trimming, exact H3 canvas resizing, a replaceable static-mask input,
+32x32 grid verification, source AV encoding, per-row masked sampling, and synchronized
+video/audio decoding.
+
+The example expects a 24 fps source video with an audio track. Its **Load Image (as Mask)**
+node reads white from the selected image's red channel and broadcasts that mask across time;
+replace this connection with ComfyUI's **SAM3 Track to Mask** output for a moving temporal
+mask. Select your own source video and mask after loading the workflow.
+
 ## Compatibility
 
 - Developed against ComfyUI `6f7cd7fceaaf60d2669b554936394a7412c6fde5` (v0.30.0 development series).
diff --git a/example_workflows/MiniMax H3 Per-Row Video Inpaint.json b/example_workflows/MiniMax H3 Per-Row Video Inpaint.json
new file mode 100644
index 0000000..89dcb5d
--- /dev/null
+++ b/example_workflows/MiniMax H3 Per-Row Video Inpaint.json	
@@ -0,0 +1,1694 @@
+{
+  "id": "57c5539b-6838-4d94-a883-55286e837d13",
+  "revision": 0,
+  "last_node_id": 31,
+  "last_link_id": 38,
+  "nodes": [
+    {
+      "id": 1,
+      "type": "LoadVideo",
+      "pos": [
+        -2500,
+        -40
+      ],
+      "size": [
+        300,
+        80
+      ],
+      "flags": {},
+      "order": 0,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "VIDEO",
+          "type": "VIDEO",
+          "links": [
+            1
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoadVideo"
+      },
+      "widgets_values": [
+        "source_video.mp4"
+      ]
+    },
+    {
+      "id": 2,
+      "type": "GetVideoComponents",
+      "pos": [
+        -2140,
+        -40
+      ],
+      "size": [
+        290,
+        110
+      ],
+      "flags": {},
+      "order": 1,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "video",
+          "type": "VIDEO",
+          "link": 1
+        }
+      ],
+      "outputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "links": [
+            2
+          ]
+        },
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "links": [
+            3
+          ]
+        },
+        {
+          "name": "fps",
+          "type": "FLOAT",
+          "links": null
+        },
+        {
+          "name": "bit_depth",
+          "type": "INT",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "GetVideoComponents"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 3,
+      "type": "MiniMaxH3TrimSourceAV",
+      "pos": [
+        -1790,
+        -40
+      ],
+      "size": [
+        360,
+        150
+      ],
+      "flags": {},
+      "order": 5,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "frames",
+          "type": "IMAGE",
+          "link": 2
+        },
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 3
+        }
+      ],
+      "outputs": [
+        {
+          "name": "trimmed_frames",
+          "type": "IMAGE",
+          "links": [
+            4
+          ]
+        },
+        {
+          "name": "trimmed_audio",
+          "type": "AUDIO",
+          "links": [
+            5
+          ]
+        },
+        {
+          "name": "h3_length",
+          "type": "INT",
+          "links": [
+            6
+          ]
+        },
+        {
+          "name": "trim_info",
+          "type": "STRING",
+          "links": [
+            7
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3TrimSourceAV"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 4,
+      "type": "PreviewAny",
+      "pos": [
+        -1760,
+        180
+      ],
+      "size": [
+        330,
+        90
+      ],
+      "flags": {},
+      "order": 9,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "source",
+          "type": "*",
+          "link": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "STRING",
+          "type": "STRING",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewAny"
+      },
+      "widgets_values": [],
+      "title": "Trim report"
+    },
+    {
+      "id": 5,
+      "type": "ImageScale",
+      "pos": [
+        -1360,
+        -40
+      ],
+      "size": [
+        310,
+        150
+      ],
+      "flags": {},
+      "order": 7,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 4
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            8,
+            9
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "ImageScale"
+      },
+      "widgets_values": [
+        "lanczos",
+        1344,
+        768,
+        "center"
+      ],
+      "title": "Resize/crop exact H3 canvas"
+    },
+    {
+      "id": 6,
+      "type": "LoadImageMask",
+      "pos": [
+        -1360,
+        310
+      ],
+      "size": [
+        310,
+        310
+      ],
+      "flags": {},
+      "order": 2,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "MASK",
+          "type": "MASK",
+          "links": [
+            10
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LoadImageMask"
+      },
+      "widgets_values": [
+        "inpaint_mask.png",
+        "red"
+      ],
+      "title": "Upload mask — white regenerates"
+    },
+    {
+      "id": 7,
+      "type": "MiniMaxH3MaskGridPreview",
+      "pos": [
+        -980,
+        140
+      ],
+      "size": [
+        370,
+        300
+      ],
+      "flags": {},
+      "order": 10,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "image",
+          "type": "IMAGE",
+          "link": 9
+        },
+        {
+          "name": "mask",
+          "type": "MASK",
+          "link": 10
+        }
+      ],
+      "outputs": [
+        {
+          "name": "snapped_mask",
+          "type": "MASK",
+          "links": [
+            11
+          ]
+        },
+        {
+          "name": "grid_preview",
+          "type": "IMAGE",
+          "links": [
+            12
+          ]
+        },
+        {
+          "name": "grid_info",
+          "type": "STRING",
+          "links": [
+            38
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3MaskGridPreview"
+      },
+      "widgets_values": [
+        "runtime exact (latent max)",
+        0,
+        0.38,
+        true,
+        true
+      ]
+    },
+    {
+      "id": 8,
+      "type": "PreviewImage",
+      "pos": [
+        -540,
+        120
+      ],
+      "size": [
+        380,
+        330
+      ],
+      "flags": {},
+      "order": 14,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 12
+        }
+      ],
+      "outputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewImage"
+      },
+      "widgets_values": [],
+      "title": "Verify generated 32×32 cells"
+    },
+    {
+      "id": 9,
+      "type": "VAELoader",
+      "pos": [
+        -2500,
+        830
+      ],
+      "size": [
+        590,
+        70
+      ],
+      "flags": {},
+      "order": 3,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            13,
+            24,
+            33
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader",
+        "models": [
+          {
+            "name": "minimax_h3_video_vae_fp16.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors",
+            "directory": "vae"
+          }
+        ]
+      },
+      "widgets_values": [
+        "minimax_h3_video_vae_fp16.safetensors"
+      ],
+      "title": "MiniMax H3 video VAE"
+    },
+    {
+      "id": 10,
+      "type": "VAELoader",
+      "pos": [
+        -2500,
+        950
+      ],
+      "size": [
+        590,
+        70
+      ],
+      "flags": {},
+      "order": 4,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "VAE",
+          "type": "VAE",
+          "links": [
+            14,
+            34
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAELoader",
+        "models": [
+          {
+            "name": "minimax_h3_audio_vae_fp32.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors",
+            "directory": "vae"
+          }
+        ]
+      },
+      "widgets_values": [
+        "minimax_h3_audio_vae_fp32.safetensors"
+      ],
+      "title": "MiniMax H3 audio VAE"
+    },
+    {
+      "id": 11,
+      "type": "VAEEncode",
+      "pos": [
+        -980,
+        -120
+      ],
+      "size": [
+        250,
+        70
+      ],
+      "flags": {},
+      "order": 11,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "pixels",
+          "type": "IMAGE",
+          "link": 8
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 13
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            15
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEEncode"
+      },
+      "widgets_values": [],
+      "title": "Encode source video"
+    },
+    {
+      "id": 12,
+      "type": "VAEEncodeAudio",
+      "pos": [
+        -980,
+        -20
+      ],
+      "size": [
+        250,
+        70
+      ],
+      "flags": {},
+      "order": 12,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 5
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 14
+        }
+      ],
+      "outputs": [
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": [
+            16
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEEncodeAudio"
+      },
+      "widgets_values": [],
+      "title": "Encode source audio"
+    },
+    {
+      "id": 13,
+      "type": "LTXVConcatAVLatent",
+      "pos": [
+        -650,
+        -80
+      ],
+      "size": [
+        290,
+        90
+      ],
+      "flags": {},
+      "order": 13,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "video_latent",
+          "type": "LATENT",
+          "link": 15
+        },
+        {
+          "name": "audio_latent",
+          "type": "LATENT",
+          "link": 16
+        }
+      ],
+      "outputs": [
+        {
+          "name": "latent",
+          "type": "LATENT",
+          "links": [
+            17
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LTXVConcatAVLatent"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 14,
+      "type": "MiniMaxH3SetGenerationMask",
+      "pos": [
+        -300,
+        -80
+      ],
+      "size": [
+        340,
+        180
+      ],
+      "flags": {},
+      "order": 15,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "av_latent",
+          "type": "LATENT",
+          "link": 17
+        },
+        {
+          "name": "mask",
+          "type": "MASK",
+          "link": 11
+        }
+      ],
+      "outputs": [
+        {
+          "name": "masked_av_latent",
+          "type": "LATENT",
+          "links": [
+            18
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3SetGenerationMask"
+      },
+      "widgets_values": [
+        "white = generate",
+        "preserve source audio"
+      ]
+    },
+    {
+      "id": 15,
+      "type": "UNETLoader",
+      "pos": [
+        -2500,
+        1200
+      ],
+      "size": [
+        590,
+        90
+      ],
+      "flags": {},
+      "order": 6,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            19
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "UNETLoader",
+        "models": [
+          {
+            "name": "minimax_h3_fl2va_pruned_int8_convrot.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_fl2va_pruned_int8_convrot.safetensors",
+            "directory": "diffusion_models"
+          }
+        ]
+      },
+      "widgets_values": [
+        "minimax_h3_fl2va_pruned_int8_convrot.safetensors",
+        "default"
+      ],
+      "title": "MiniMax H3 model"
+    },
+    {
+      "id": 16,
+      "type": "MiniMaxH3SigmaShift",
+      "pos": [
+        -1840,
+        1200
+      ],
+      "size": [
+        300,
+        120
+      ],
+      "flags": {},
+      "order": 16,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 19
+        }
+      ],
+      "outputs": [
+        {
+          "name": "MODEL",
+          "type": "MODEL",
+          "links": [
+            20
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3SigmaShift"
+      },
+      "widgets_values": [
+        12,
+        3
+      ]
+    },
+    {
+      "id": 17,
+      "type": "MiniMaxH3PerRowMaskPatch",
+      "pos": [
+        -1470,
+        1200
+      ],
+      "size": [
+        350,
+        90
+      ],
+      "flags": {},
+      "order": 17,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 20
+        }
+      ],
+      "outputs": [
+        {
+          "name": "patched_model",
+          "type": "MODEL",
+          "links": [
+            21,
+            22
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3PerRowMaskPatch"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 18,
+      "type": "CLIPLoader",
+      "pos": [
+        -2500,
+        1390
+      ],
+      "size": [
+        590,
+        120
+      ],
+      "flags": {},
+      "order": 8,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "CLIP",
+          "type": "CLIP",
+          "links": [
+            23
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CLIPLoader",
+        "models": [
+          {
+            "name": "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
+            "url": "https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
+            "directory": "text_encoders"
+          }
+        ]
+      },
+      "widgets_values": [
+        "qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
+        "minimax",
+        "default"
+      ],
+      "title": "MiniMax H3 text encoder"
+    },
+    {
+      "id": 19,
+      "type": "MiniMaxH3ImageToVideo",
+      "pos": [
+        -1040,
+        1060
+      ],
+      "size": [
+        430,
+        510
+      ],
+      "flags": {},
+      "order": 18,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "clip",
+          "type": "CLIP",
+          "link": 23
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 24
+        },
+        {
+          "name": "first_frame",
+          "type": "IMAGE",
+          "link": null,
+          "shape": 7
+        },
+        {
+          "name": "last_frame",
+          "type": "IMAGE",
+          "link": null,
+          "shape": 7
+        },
+        {
+          "name": "prompt",
+          "type": "STRING",
+          "link": null,
+          "widget": {
+            "name": "prompt"
+          }
+        },
+        {
+          "name": "width",
+          "type": "INT",
+          "link": null,
+          "widget": {
+            "name": "width"
+          }
+        },
+        {
+          "name": "height",
+          "type": "INT",
+          "link": null,
+          "widget": {
+            "name": "height"
+          }
+        },
+        {
+          "name": "length",
+          "type": "INT",
+          "link": 6,
+          "widget": {
+            "name": "length"
+          }
+        }
+      ],
+      "outputs": [
+        {
+          "name": "positive",
+          "type": "CONDITIONING",
+          "links": [
+            25
+          ]
+        },
+        {
+          "name": "LATENT",
+          "type": "LATENT",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "MiniMaxH3ImageToVideo"
+      },
+      "widgets_values": [
+        "Video editing. Regenerate only the selected masked region. Replace the object inside that region with a polished red metal sphere that follows the original object's position, scale, perspective, and motion throughout the complete shot. Preserve every unmasked subject, the background, camera movement, framing, lighting, shadows, timing, and scene composition exactly. Keep the source audio unchanged and synchronized.",
+        1344,
+        768,
+        124
+      ],
+      "title": "Prompt conditioning only — ignore empty latent"
+    },
+    {
+      "id": 20,
+      "type": "BasicGuider",
+      "pos": [
+        -500,
+        1120
+      ],
+      "size": [
+        300,
+        80
+      ],
+      "flags": {},
+      "order": 20,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 22
+        },
+        {
+          "name": "conditioning",
+          "type": "CONDITIONING",
+          "link": 25
+        }
+      ],
+      "outputs": [
+        {
+          "name": "GUIDER",
+          "type": "GUIDER",
+          "links": [
+            26
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "BasicGuider"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 21,
+      "type": "BasicScheduler",
+      "pos": [
+        -500,
+        1240
+      ],
+      "size": [
+        300,
+        140
+      ],
+      "flags": {},
+      "order": 19,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "model",
+          "type": "MODEL",
+          "link": 21
+        }
+      ],
+      "outputs": [
+        {
+          "name": "SIGMAS",
+          "type": "SIGMAS",
+          "links": [
+            27
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "BasicScheduler"
+      },
+      "widgets_values": [
+        "simple",
+        20,
+        1
+      ]
+    },
+    {
+      "id": 22,
+      "type": "KSamplerSelect",
+      "pos": [
+        -500,
+        1420
+      ],
+      "size": [
+        300,
+        80
+      ],
+      "flags": {},
+      "order": 21,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "SAMPLER",
+          "type": "SAMPLER",
+          "links": [
+            28
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "KSamplerSelect"
+      },
+      "widgets_values": [
+        "res_multistep"
+      ]
+    },
+    {
+      "id": 23,
+      "type": "RandomNoise",
+      "pos": [
+        -500,
+        1540
+      ],
+      "size": [
+        300,
+        90
+      ],
+      "flags": {},
+      "order": 22,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [
+        {
+          "name": "NOISE",
+          "type": "NOISE",
+          "links": [
+            29
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "RandomNoise"
+      },
+      "widgets_values": [
+        1,
+        "randomize"
+      ]
+    },
+    {
+      "id": 24,
+      "type": "SamplerCustomAdvanced",
+      "pos": [
+        -100,
+        1190
+      ],
+      "size": [
+        330,
+        180
+      ],
+      "flags": {},
+      "order": 23,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "noise",
+          "type": "NOISE",
+          "link": 29
+        },
+        {
+          "name": "guider",
+          "type": "GUIDER",
+          "link": 26
+        },
+        {
+          "name": "sampler",
+          "type": "SAMPLER",
+          "link": 28
+        },
+        {
+          "name": "sigmas",
+          "type": "SIGMAS",
+          "link": 27
+        },
+        {
+          "name": "latent_image",
+          "type": "LATENT",
+          "link": 18
+        }
+      ],
+      "outputs": [
+        {
+          "name": "output",
+          "type": "LATENT",
+          "links": [
+            30
+          ]
+        },
+        {
+          "name": "denoised_output",
+          "type": "LATENT",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SamplerCustomAdvanced"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 25,
+      "type": "LTXVSeparateAVLatent",
+      "pos": [
+        300,
+        1190
+      ],
+      "size": [
+        300,
+        90
+      ],
+      "flags": {},
+      "order": 24,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "av_latent",
+          "type": "LATENT",
+          "link": 30
+        }
+      ],
+      "outputs": [
+        {
+          "name": "video_latent",
+          "type": "LATENT",
+          "links": [
+            31
+          ]
+        },
+        {
+          "name": "audio_latent",
+          "type": "LATENT",
+          "links": [
+            32
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "LTXVSeparateAVLatent"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 26,
+      "type": "VAEDecode",
+      "pos": [
+        670,
+        1110
+      ],
+      "size": [
+        250,
+        70
+      ],
+      "flags": {},
+      "order": 25,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 31
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 33
+        }
+      ],
+      "outputs": [
+        {
+          "name": "IMAGE",
+          "type": "IMAGE",
+          "links": [
+            35
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecode"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 27,
+      "type": "VAEDecodeAudio",
+      "pos": [
+        670,
+        1240
+      ],
+      "size": [
+        250,
+        70
+      ],
+      "flags": {},
+      "order": 26,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "samples",
+          "type": "LATENT",
+          "link": 32
+        },
+        {
+          "name": "vae",
+          "type": "VAE",
+          "link": 34
+        }
+      ],
+      "outputs": [
+        {
+          "name": "AUDIO",
+          "type": "AUDIO",
+          "links": [
+            36
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "VAEDecodeAudio"
+      },
+      "widgets_values": []
+    },
+    {
+      "id": 28,
+      "type": "CreateVideo",
+      "pos": [
+        990,
+        1140
+      ],
+      "size": [
+        280,
+        120
+      ],
+      "flags": {},
+      "order": 27,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "images",
+          "type": "IMAGE",
+          "link": 35
+        },
+        {
+          "name": "audio",
+          "type": "AUDIO",
+          "link": 36,
+          "shape": 7
+        }
+      ],
+      "outputs": [
+        {
+          "name": "VIDEO",
+          "type": "VIDEO",
+          "links": [
+            37
+          ]
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "CreateVideo"
+      },
+      "widgets_values": [
+        24,
+        8
+      ]
+    },
+    {
+      "id": 29,
+      "type": "SaveVideo",
+      "pos": [
+        1340,
+        1120
+      ],
+      "size": [
+        520,
+        130
+      ],
+      "flags": {},
+      "order": 28,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "video",
+          "type": "VIDEO",
+          "link": 37
+        }
+      ],
+      "outputs": [
+        {
+          "name": "video",
+          "type": "VIDEO",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "SaveVideo"
+      },
+      "widgets_values": [
+        "video/MiniMax_H3_Inpaint",
+        "auto",
+        "auto"
+      ]
+    },
+    {
+      "id": 30,
+      "type": "MarkdownNote",
+      "pos": [
+        -2500,
+        -650
+      ],
+      "size": [
+        900,
+        500
+      ],
+      "flags": {},
+      "order": 29,
+      "mode": 0,
+      "inputs": [],
+      "outputs": [],
+      "properties": {},
+      "widgets_values": [
+        "## MiniMax H3 per-row video inpainting\n\n1. Upload a **24 fps** source video containing an audio track.\n2. Upload or paint a mask image. **White regenerates; black preserves.** One mask image is broadcast across the full video.\n3. Edit the H3 prompt to describe the intended replacement inside the mask.\n4. Queue the workflow.\n\nThe **Trim Source AV** node drops trailing frames to the largest valid `17k+5` length and trims audio to the same 24 fps duration. Its `h3_length` output controls H3 conditioning.\n\nThe resized source frames feed all three spatial branches: video VAE encoding, the 32×32 grid preview, and any optional tracking. Replace the static Load Image mask connection with **SAM3 Track to Mask** for a moving temporal mask.\n\nThe source video is the main latent and must not also be connected as an H3 reference video. Use the MiniMax H3 video VAE for frames and the MiniMax H3 audio VAE for audio."
+      ],
+      "title": "How to use this workflow",
+      "color": "#222",
+      "bgcolor": "#000"
+    },
+    {
+      "id": 31,
+      "type": "PreviewAny",
+      "pos": [
+        -540,
+        500
+      ],
+      "size": [
+        380,
+        90
+      ],
+      "flags": {},
+      "order": 30,
+      "mode": 0,
+      "inputs": [
+        {
+          "name": "source",
+          "type": "*",
+          "link": 38
+        }
+      ],
+      "outputs": [
+        {
+          "name": "STRING",
+          "type": "STRING",
+          "links": null
+        }
+      ],
+      "properties": {
+        "Node name for S&R": "PreviewAny"
+      },
+      "widgets_values": [],
+      "title": "H3 grid report"
+    }
+  ],
+  "links": [
+    [
+      1,
+      1,
+      0,
+      2,
+      0,
+      "VIDEO"
+    ],
+    [
+      2,
+      2,
+      0,
+      3,
+      0,
+      "IMAGE"
+    ],
+    [
+      3,
+      2,
+      1,
+      3,
+      1,
+      "AUDIO"
+    ],
+    [
+      4,
+      3,
+      0,
+      5,
+      0,
+      "IMAGE"
+    ],
+    [
+      5,
+      3,
+      1,
+      12,
+      0,
+      "AUDIO"
+    ],
+    [
+      6,
+      3,
+      2,
+      19,
+      7,
+      "INT"
+    ],
+    [
+      7,
+      3,
+      3,
+      4,
+      0,
+      "STRING"
+    ],
+    [
+      8,
+      5,
+      0,
+      11,
+      0,
+      "IMAGE"
+    ],
+    [
+      9,
+      5,
+      0,
+      7,
+      0,
+      "IMAGE"
+    ],
+    [
+      10,
+      6,
+      0,
+      7,
+      1,
+      "MASK"
+    ],
+    [
+      11,
+      7,
+      0,
+      14,
+      1,
+      "MASK"
+    ],
+    [
+      12,
+      7,
+      1,
+      8,
+      0,
+      "IMAGE"
+    ],
+    [
+      13,
+      9,
+      0,
+      11,
+      1,
+      "VAE"
+    ],
+    [
+      14,
+      10,
+      0,
+      12,
+      1,
+      "VAE"
+    ],
+    [
+      15,
+      11,
+      0,
+      13,
+      0,
+      "LATENT"
+    ],
+    [
+      16,
+      12,
+      0,
+      13,
+      1,
+      "LATENT"
+    ],
+    [
+      17,
+      13,
+      0,
+      14,
+      0,
+      "LATENT"
+    ],
+    [
+      18,
+      14,
+      0,
+      24,
+      4,
+      "LATENT"
+    ],
+    [
+      19,
+      15,
+      0,
+      16,
+      0,
+      "MODEL"
+    ],
+    [
+      20,
+      16,
+      0,
+      17,
+      0,
+      "MODEL"
+    ],
+    [
+      21,
+      17,
+      0,
+      21,
+      0,
+      "MODEL"
+    ],
+    [
+      22,
+      17,
+      0,
+      20,
+      0,
+      "MODEL"
+    ],
+    [
+      23,
+      18,
+      0,
+      19,
+      0,
+      "CLIP"
+    ],
+    [
+      24,
+      9,
+      0,
+      19,
+      1,
+      "VAE"
+    ],
+    [
+      25,
+      19,
+      0,
+      20,
+      1,
+      "CONDITIONING"
+    ],
+    [
+      26,
+      20,
+      0,
+      24,
+      1,
+      "GUIDER"
+    ],
+    [
+      27,
+      21,
+      0,
+      24,
+      3,
+      "SIGMAS"
+    ],
+    [
+      28,
+      22,
+      0,
+      24,
+      2,
+      "SAMPLER"
+    ],
+    [
+      29,
+      23,
+      0,
+      24,
+      0,
+      "NOISE"
+    ],
+    [
+      30,
+      24,
+      0,
+      25,
+      0,
+      "LATENT"
+    ],
+    [
+      31,
+      25,
+      0,
+      26,
+      0,
+      "LATENT"
+    ],
+    [
+      32,
+      25,
+      1,
+      27,
+      0,
+      "LATENT"
+    ],
+    [
+      33,
+      9,
+      0,
+      26,
+      1,
+      "VAE"
+    ],
+    [
+      34,
+      10,
+      0,
+      27,
+      1,
+      "VAE"
+    ],
+    [
+      35,
+      26,
+      0,
+      28,
+      0,
+      "IMAGE"
+    ],
+    [
+      36,
+      27,
+      0,
+      28,
+      1,
+      "AUDIO"
+    ],
+    [
+      37,
+      28,
+      0,
+      29,
+      0,
+      "VIDEO"
+    ],
+    [
+      38,
+      7,
+      2,
+      31,
+      0,
+      "STRING"
+    ]
+  ],
+  "groups": [
+    {
+      "id": 1,
+      "title": "1. Source video, valid H3 length, and generation mask",
+      "bounding": [
+        -2550,
+        -170,
+        2500,
+        850
+      ],
+      "color": "#3f789e",
+      "flags": {}
+    },
+    {
+      "id": 2,
+      "title": "2. MiniMax H3 models",
+      "bounding": [
+        -2550,
+        740,
+        1500,
+        900
+      ],
+      "color": "#3f789e",
+      "flags": {}
+    },
+    {
+      "id": 3,
+      "title": "3. Prompt conditioning",
+      "bounding": [
+        -1080,
+        950,
+        940,
+        700
+      ],
+      "color": "#3f789e",
+      "flags": {}
+    },
+    {
+      "id": 4,
+      "title": "4. Per-row masked sampling",
+      "bounding": [
+        -560,
+        1030,
+        850,
+        700
+      ],
+      "color": "#3f789e",
+      "flags": {}
+    },
+    {
+      "id": 5,
+      "title": "5. Decode source-preserving result",
+      "bounding": [
+        250,
+        1030,
+        1660,
+        600
+      ],
+      "color": "#3f789e",
+      "flags": {}
+    }
+  ],
+  "config": {},
+  "extra": {
+    "frontendVersion": "1.48.6",
+    "ds": {
+      "scale": 0.62,
+      "offset": [
+        1630,
+        420
+      ]
+    }
+  },
+  "version": 0.4
+}
diff --git a/tests/test_example_workflow.py b/tests/test_example_workflow.py
new file mode 100644
index 0000000..b03e12b
--- /dev/null
+++ b/tests/test_example_workflow.py
@@ -0,0 +1,73 @@
+import json
+from pathlib import Path
+
+
+WORKFLOW = (
+    Path(__file__).parents[1]
+    / "example_workflows"
+    / "MiniMax H3 Per-Row Video Inpaint.json"
+)
+
+
+def _load_workflow():
+    return json.loads(WORKFLOW.read_text(encoding="utf-8"))
+
+
+def test_inpaint_workflow_links_are_consistent():
+    workflow = _load_workflow()
+    nodes = {node["id"]: node for node in workflow["nodes"]}
+    link_ids = set()
+
+    for link in workflow["links"]:
+        link_id, origin_id, origin_slot, target_id, target_slot, link_type = link
+        assert link_id not in link_ids
+        link_ids.add(link_id)
+        assert origin_id in nodes
+        assert target_id in nodes
+        output = nodes[origin_id]["outputs"][origin_slot]
+        input_ = nodes[target_id]["inputs"][target_slot]
+        assert link_id in output["links"]
+        assert input_["link"] == link_id
+        assert output["type"] == link_type
+        assert input_["type"] in {link_type, "*"}
+
+    assert workflow["last_link_id"] == max(link_ids)
+    assert workflow["last_node_id"] == max(nodes)
+
+
+def test_inpaint_workflow_contains_complete_h3_mask_path():
+    workflow = _load_workflow()
+    types = {node["type"] for node in workflow["nodes"]}
+    assert {
+        "MiniMaxH3TrimSourceAV",
+        "MiniMaxH3MaskGridPreview",
+        "MiniMaxH3SetGenerationMask",
+        "MiniMaxH3SigmaShift",
+        "MiniMaxH3PerRowMaskPatch",
+        "MiniMaxH3ImageToVideo",
+        "LTXVConcatAVLatent",
+        "SamplerCustomAdvanced",
+        "LTXVSeparateAVLatent",
+    } <= types
+
+
+def test_trimmed_length_controls_conditioning_and_source_latent_controls_sampler():
+    workflow = _load_workflow()
+    nodes = {node["id"]: node for node in workflow["nodes"]}
+    links = {link[0]: link for link in workflow["links"]}
+
+    trim = next(node for node in nodes.values() if node["type"] == "MiniMaxH3TrimSourceAV")
+    conditioning = next(
+        node for node in nodes.values() if node["type"] == "MiniMaxH3ImageToVideo"
+    )
+    sampler = next(node for node in nodes.values() if node["type"] == "SamplerCustomAdvanced")
+    set_mask = next(
+        node for node in nodes.values() if node["type"] == "MiniMaxH3SetGenerationMask"
+    )
+
+    length_link = links[conditioning["inputs"][7]["link"]]
+    assert length_link[1:3] == [trim["id"], 2]
+
+    latent_link = links[sampler["inputs"][4]["link"]]
+    assert latent_link[1:3] == [set_mask["id"], 0]
+    assert conditioning["outputs"][1]["links"] is None