ComfyUI-Workflows / MiniMax H3 /H3_Studio_Unified_Image.json
thaakeno46's picture
Add MiniMax H3 Studio Unified Image workflow
2e73a2f verified
Raw
History Blame Contribute Delete
58.9 kB
{
"id": "51ffc0bb-1b7a-4a1c-a183-1ce99edb4e5e",
"revision": 0,
"last_node_id": 30,
"last_link_id": 19,
"nodes": [
{
"id": 10,
"type": "H3StudioDirector",
"pos": [
-1450,
220
],
"size": [
700,
780
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [
{
"name": "mode",
"type": "COMBO",
"link": null,
"widget": {
"name": "mode"
}
},
{
"name": "prompt",
"type": "STRING",
"link": null,
"widget": {
"name": "prompt"
}
},
{
"name": "resolution",
"type": "COMBO",
"link": null,
"widget": {
"name": "resolution"
}
},
{
"name": "aspect_ratio",
"type": "COMBO",
"link": null,
"widget": {
"name": "aspect_ratio"
}
},
{
"name": "width",
"type": "INT",
"link": null,
"widget": {
"name": "width"
}
},
{
"name": "height",
"type": "INT",
"link": null,
"widget": {
"name": "height"
}
},
{
"name": "seconds",
"type": "FLOAT",
"link": null,
"widget": {
"name": "seconds"
}
},
{
"name": "advanced",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "advanced"
}
},
{
"name": "fps",
"type": "FLOAT",
"link": null,
"widget": {
"name": "fps"
}
},
{
"name": "keyframe_role",
"type": "COMBO",
"link": null,
"widget": {
"name": "keyframe_role"
}
},
{
"name": "ref_image_size",
"type": "COMBO",
"link": null,
"widget": {
"name": "ref_image_size"
}
},
{
"name": "reference_mention_mode",
"type": "COMBO",
"link": null,
"widget": {
"name": "reference_mention_mode"
}
},
{
"name": "megapixels",
"type": "FLOAT",
"link": null,
"widget": {
"name": "megapixels"
}
},
{
"name": "seed",
"type": "INT",
"link": null,
"widget": {
"name": "seed"
}
},
{
"name": "enhance_mode",
"type": "COMBO",
"link": null,
"widget": {
"name": "enhance_mode"
}
},
{
"name": "adherence",
"type": "FLOAT",
"link": null,
"widget": {
"name": "adherence"
}
},
{
"name": "route",
"type": "COMBO",
"link": null,
"widget": {
"name": "route"
}
},
{
"name": "sampling_profile",
"type": "COMBO",
"link": null,
"widget": {
"name": "sampling_profile"
}
},
{
"name": "frame_profile",
"type": "COMBO",
"link": null,
"widget": {
"name": "frame_profile"
}
},
{
"name": "analyzer_model",
"type": "STRING",
"link": null,
"widget": {
"name": "analyzer_model"
}
},
{
"name": "studio_state",
"type": "STRING",
"link": null,
"widget": {
"name": "studio_state"
}
},
{
"name": "media",
"type": "*",
"link": null,
"shape": 7
},
{
"name": "media_filename",
"type": "STRING",
"link": null,
"widget": {
"name": "media_filename"
}
},
{
"name": "h3_bundle",
"type": "H3_STUDIO_BUNDLE",
"link": 3
}
],
"outputs": [
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"links": [
1,
10,
13,
14,
15
]
},
{
"name": "compiled_prompt",
"type": "STRING",
"links": null
},
{
"name": "state_json",
"type": "STRING",
"links": null
},
{
"name": "width",
"type": "INT",
"links": null
},
{
"name": "height",
"type": "INT",
"links": null
},
{
"name": "seed",
"type": "INT",
"links": [
9
]
},
{
"name": "diagnostics",
"type": "STRING",
"links": null
}
],
"title": "MiniMax H3 Studio · Image Director",
"properties": {
"Node name for S&R": "H3StudioDirector",
"h3studio_virtual_media_links": [],
"h3studio_state": "{\"schema_version\":10,\"prompt\":\"Describe the final still image.\",\"references\":[],\"prompt_options\":{\"enhance_mode\":\"compile_only\",\"analyze_images\":true,\"deep_enhancement\":true,\"analyzer_resolution\":512,\"adherence\":0.85,\"detail_level\":\"detailed\",\"preserve_user_text\":true,\"infer_roles\":true,\"system_instruction\":\"\",\"analyzer_model\":\"\",\"analyzer_device\":\"auto\",\"analyzer_quantization\":\"auto\",\"analyzer_max_tokens\":1800,\"analyzer_keep_loaded\":false},\"generation\":{\"mode\":\"auto\",\"route\":\"auto\",\"seed\":42,\"seed_locked\":false,\"aspect_ratio\":\"1:1\",\"megapixels\":1.0,\"custom_width\":1024,\"custom_height\":1024,\"cap_native_resolution\":true,\"sampling_profile\":\"base_quality_20\",\"frame_profile\":\"recommended_5\",\"frame_selection\":\"decode_recommended\",\"reference_short_edge\":2048,\"source_image_ordinal\":1},\"ui\":{\"advanced_open\":false,\"reference_details\":{}}}",
"h3studio_prompt_reference_doc": {
"version": 1,
"parts": [
{
"type": "text",
"text": "Describe the final still image."
}
]
}
},
"widgets_values": [
"image",
"Describe the final still image. Upload references here only when they have a specific visual job.",
"Custom",
"1:1",
1024,
1024,
5.0,
false,
24.0,
"first",
"2k",
"index",
1.0,
42,
"vlm",
0.85,
"auto",
"base_quality_20",
"recommended_5",
"",
"{\"schema_version\":10,\"prompt\":\"Describe the final still image.\",\"references\":[],\"prompt_options\":{\"enhance_mode\":\"compile_only\",\"analyze_images\":true,\"deep_enhancement\":true,\"analyzer_resolution\":512,\"adherence\":0.85,\"detail_level\":\"detailed\",\"preserve_user_text\":true,\"infer_roles\":true,\"system_instruction\":\"\",\"analyzer_model\":\"\",\"analyzer_device\":\"auto\",\"analyzer_quantization\":\"auto\",\"analyzer_max_tokens\":1800,\"analyzer_keep_loaded\":false},\"generation\":{\"mode\":\"auto\",\"route\":\"auto\",\"seed\":42,\"seed_locked\":false,\"aspect_ratio\":\"1:1\",\"megapixels\":1.0,\"custom_width\":1024,\"custom_height\":1024,\"cap_native_resolution\":true,\"sampling_profile\":\"base_quality_20\",\"frame_profile\":\"recommended_5\",\"frame_selection\":\"decode_recommended\",\"reference_short_edge\":2048,\"source_image_ordinal\":1},\"ui\":{\"advanced_open\":false,\"reference_details\":{}}}",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
"",
"image",
"",
"auto",
"attribute_transfer",
""
],
"color": "#255049",
"bgcolor": "#172d29"
},
{
"id": 11,
"type": "H3StudioLoader",
"pos": [
-520,
260
],
"size": [
560,
220
],
"flags": {},
"order": 1,
"mode": 0,
"inputs": [
{
"name": "fl2va_model",
"type": "COMBO",
"link": null,
"widget": {
"name": "fl2va_model"
}
},
{
"name": "ref2va_model",
"type": "COMBO",
"link": null,
"widget": {
"name": "ref2va_model"
}
},
{
"name": "text_encoder",
"type": "COMBO",
"link": null,
"widget": {
"name": "text_encoder"
}
},
{
"name": "video_vae",
"type": "COMBO",
"link": null,
"widget": {
"name": "video_vae"
}
},
{
"name": "image_vae",
"type": "COMBO",
"link": null,
"widget": {
"name": "image_vae"
}
},
{
"name": "image_analyzer",
"type": "COMBO",
"link": null,
"widget": {
"name": "image_analyzer"
}
},
{
"name": "prompt_writer",
"type": "COMBO",
"link": null,
"widget": {
"name": "prompt_writer"
}
}
],
"outputs": [
{
"name": "h3_bundle",
"type": "H3_STUDIO_BUNDLE",
"links": [
2,
3,
12
]
},
{
"name": "clip",
"type": "CLIP",
"links": null
},
{
"name": "video_vae",
"type": "VAE",
"links": null
},
{
"name": "model_info",
"type": "STRING",
"links": null
}
],
"title": "H3 models · lazy route loader",
"properties": {
"Node name for S&R": "H3StudioLoader"
},
"widgets_values": [
"minimax_h3_fl2va_pruned_w4a8_mixed.safetensors",
"minimax_h3_ref2va_pruned_w4a8_mixed.safetensors",
"qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors",
"minimax_h3_video_vae_fp16.safetensors",
"Disabled - original H3 video VAE only",
"Auto · Qwen3-VL 4B",
"Same as image analyzer"
],
"color": "#31475c",
"bgcolor": "#202e3c"
},
{
"id": 12,
"type": "H3StudioCondition",
"pos": [
-520,
560
],
"size": [
560,
150
],
"flags": {},
"order": 2,
"mode": 0,
"inputs": [
{
"name": "h3_bundle",
"type": "H3_STUDIO_BUNDLE",
"link": 2
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 1
}
],
"outputs": [
{
"name": "model",
"type": "MODEL",
"links": [
4
]
},
{
"name": "generation",
"type": "H3_STUDIO_GENERATION",
"links": null
},
{
"name": "positive",
"type": "CONDITIONING",
"links": [
6
]
},
{
"name": "h3_latent",
"type": "LATENT",
"links": [
7
]
},
{
"name": "video_vae",
"type": "VAE",
"links": [
8
]
},
{
"name": "requested_frames",
"type": "INT",
"links": null
},
{
"name": "run_info",
"type": "STRING",
"links": null
}
],
"title": "Compile, condition, and route once",
"properties": {
"Node name for S&R": "H3StudioCondition"
},
"widgets_values": [],
"color": "#53482c",
"bgcolor": "#352e1d"
},
{
"id": 16,
"type": "H3StudioTAEH3Preview",
"pos": [
220,
260
],
"size": [
620,
620
],
"flags": {},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 4
},
{
"name": "enabled",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "enabled"
}
},
{
"name": "tiny_vae",
"type": "COMBO",
"link": null,
"widget": {
"name": "tiny_vae"
}
},
{
"name": "max_resolution",
"type": "INT",
"link": null,
"widget": {
"name": "max_resolution"
}
},
{
"name": "jpeg_quality",
"type": "INT",
"link": null,
"widget": {
"name": "jpeg_quality"
}
},
{
"name": "preview_every_n_steps",
"type": "INT",
"link": null,
"widget": {
"name": "preview_every_n_steps"
}
}
],
"outputs": [
{
"name": "model",
"type": "MODEL",
"links": [
5
]
}
],
"title": "Live preview · TAEH3 (optional)",
"properties": {
"Node name for S&R": "H3StudioTAEH3Preview"
},
"widgets_values": [
false,
"taeh3.safetensors",
768,
90,
1
],
"color": "#31475c",
"bgcolor": "#202e3c"
},
{
"id": 13,
"type": "5930b00d-9f8e-4b87-9cb5-ff5f7cf3b30a",
"pos": [
220,
920
],
"size": [
620,
420
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 5
},
{
"name": "positive",
"type": "CONDITIONING",
"link": 6
},
{
"name": "h3_latent",
"type": "LATENT",
"link": 7
},
{
"name": "video_vae",
"type": "VAE",
"link": 8
},
{
"name": "seed",
"type": "INT",
"link": 9
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 10
}
],
"outputs": [
{
"name": "image",
"type": "IMAGE",
"links": [
11
]
},
{
"name": "decode_info",
"type": "STRING",
"links": null
},
{
"name": "selection_report",
"type": "STRING",
"links": null
}
],
"title": "H3 Studio · Sampling + exact still",
"properties": {
"cnr_id": "comfy-core",
"ver": "0.30.0",
"previewExposures": []
},
"widgets_values": [],
"color": "#3c514c",
"bgcolor": "#24312f"
},
{
"id": 14,
"type": "PreviewImage",
"pos": [
1030,
260
],
"size": [
440,
420
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 17
}
],
"outputs": [],
"title": "Final still · preview",
"properties": {
"Node name for S&R": "PreviewImage"
},
"widgets_values": [],
"color": "#3c514c",
"bgcolor": "#24312f"
},
{
"id": 15,
"type": "H3StudioSaveImage",
"pos": [
1030,
750
],
"size": [
440,
190
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 18
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 14
},
{
"name": "filename_prefix",
"type": "STRING",
"link": null,
"widget": {
"name": "filename_prefix"
}
}
],
"outputs": [],
"title": "Final still · restorable PNG",
"properties": {
"Node name for S&R": "H3StudioSaveImage"
},
"widgets_values": [
"H3Studio/%date:yyyy-MM-dd%/H3_%seed%"
],
"color": "#3c514c",
"bgcolor": "#24312f"
},
{
"id": 30,
"type": "H3StudioComparisonView",
"pos": [
1030,
1010
],
"size": [
440,
420
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "images",
"type": "IMAGE",
"link": 19
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 15
}
],
"outputs": [],
"title": "Reference comparison · optional",
"properties": {
"Node name for S&R": "H3StudioComparisonView"
},
"widgets_values": [],
"color": "#3c514c",
"bgcolor": "#24312f"
},
{
"id": 17,
"type": "H3StudioABComparison",
"pos": [
-520,
1700
],
"size": [
780,
520
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "h3_bundle",
"type": "H3_STUDIO_BUNDLE",
"link": 12
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 13
},
{
"name": "comparison_kind",
"type": "COMBO",
"link": null,
"widget": {
"name": "comparison_kind"
}
},
{
"name": "profiles",
"type": "STRING",
"link": null,
"widget": {
"name": "profiles"
}
},
{
"name": "megapixels",
"type": "STRING",
"link": null,
"widget": {
"name": "megapixels"
}
},
{
"name": "repeats",
"type": "INT",
"link": null,
"widget": {
"name": "repeats"
}
},
{
"name": "seed_strategy",
"type": "COMBO",
"link": null,
"widget": {
"name": "seed_strategy"
}
},
{
"name": "seed_step",
"type": "INT",
"link": null,
"widget": {
"name": "seed_step"
}
},
{
"name": "grid_cell_size",
"type": "INT",
"link": null,
"widget": {
"name": "grid_cell_size"
}
},
{
"name": "max_generations",
"type": "INT",
"link": null,
"widget": {
"name": "max_generations"
}
},
{
"name": "allow_large_matrix",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "allow_large_matrix"
}
},
{
"name": "include_reference_context",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "include_reference_context"
}
},
{
"name": "include_original_prompt",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "include_original_prompt"
}
},
{
"name": "live_cell_previews",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "live_cell_previews"
}
}
],
"outputs": [
{
"name": "comparison_grid",
"type": "IMAGE",
"links": [
16
]
},
{
"name": "comparison_report",
"type": "STRING",
"links": null
}
],
"title": "Benchmark Lab - profiles / resolution / VAE",
"properties": {
"Node name for S&R": "H3StudioABComparison"
},
"widgets_values": [
"Sampling profiles x resolution",
"base_quality_20, lightx_er_sde_4",
"0.40, 1.00, 2.00",
1,
"Same seed for all - fair comparison",
1,
640,
24,
false,
true,
true,
true
],
"color": "#60467a",
"bgcolor": "#332640"
},
{
"id": 19,
"type": "H3StudioLazyImageSwitch",
"pos": [
340,
1700
],
"size": [
420,
180
],
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{
"name": "benchmark_enabled",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "benchmark_enabled"
}
},
{
"name": "normal_image",
"type": "IMAGE",
"link": 11
},
{
"name": "benchmark_image",
"type": "IMAGE",
"link": 16
}
],
"outputs": [
{
"name": "image",
"type": "IMAGE",
"links": [
17,
18,
19
]
},
{
"name": "selected_mode",
"type": "STRING",
"links": null
}
],
"title": "Run mode - normal or benchmark only",
"properties": {
"Node name for S&R": "H3StudioLazyImageSwitch"
},
"widgets_values": [
false
],
"color": "#60467a",
"bgcolor": "#332640"
},
{
"id": 20,
"type": "H3StudioWorkflowNote",
"pos": [
-1450,
-180
],
"size": [
500,
320
],
"flags": {},
"order": 10,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "START HERE · three-minute setup",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"quick start",
"## Make an image\n\n1. Write naturally in the **Director**.\n2. Add references only when needed; they become ordered `@Image1`, `@Image2`, … cards.\n3. Choose the image role, retention, aspect, resolution, and seed.\n4. Queue the workflow.\n\n> [!TIP]\n> With no references, Auto uses FL2VA text-to-image immediately. Multi-reference work uses REF2VA."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 21,
"type": "H3StudioWorkflowNote",
"pos": [
-910,
-180
],
"size": [
520,
320
],
"flags": {},
"order": 11,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Prompt enhancement · what actually happens",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"settings",
"## Two deliberate Qwen stages\n\n- **Analyze image pixels**: Qwen3-VL creates factual source descriptions and repairs automatic roles.\n- **Prompt enhancement**: a compact text-only Qwen pass turns the request and facts into a production direction.\n\nThe default writer is **Same as image analyzer**, so Auto 4B does both jobs with one loaded checkpoint. Select Auto 8B or an explicit file only when you accept staging a second model. Fingerprints and both stage caches survive normal reruns."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 22,
"type": "H3StudioWorkflowNote",
"pos": [
-370,
-180
],
"size": [
520,
320
],
"flags": {},
"order": 12,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Routing · deliberate defaults",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"settings",
"## Auto is the safe default\n\n- **0 references:** FL2VA text-to-image\n- **1 anchor:** FL2VA image-to-image when requested\n- **Multiple references:** REF2VA\n\n> [!WARNING]\n> Forced routes are diagnostic controls. Invalid reference/route combinations are rejected before model work starts."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 24,
"type": "H3StudioWorkflowNote",
"pos": [
170,
-180
],
"size": [
520,
320
],
"flags": {},
"order": 13,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Why the Director is not hidden in the subgraph",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"optional / experimental",
"## Why this layout is intentional\n\nThe visible **Director** owns rich Studio state, reference cards, and `@Image` editing. The reusable subgraph owns ordinary typed sampling and exact-frame decode sockets.\n\nKeeping that boundary avoids invisible promoted widgets and makes the main path readable."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 25,
"type": "H3StudioWorkflowNote",
"pos": [
1560,
260
],
"size": [
430,
390
],
"flags": {},
"order": 15,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Image reference semantics",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"settings",
"## Choose what each image contributes\n\n- `identity` / `character` + `fully_preserved`: keep the exact person or design.\n- `style` + `attribute_transfer`: borrow rendering language, not content.\n- `composition` / `layout`: borrow placement and hierarchy.\n- `outfit`, `pose`, `typography`, `lighting`, `texture`, `object`, and `environment`: narrow transfers.\n\n**Explicit card metadata always wins** over inference."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 26,
"type": "H3StudioWorkflowNote",
"pos": [
1560,
700
],
"size": [
430,
310
],
"flags": {},
"order": 16,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Sampling and frame extraction",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"models",
"## Profiles and decode\n\n- **Base Quality / Balanced:** native H3; no acceleration files.\n- **LightX v0.1:** Kijai's empirical four-step ComfyUI recipe.\n- **PDD 600 / 900:** four-step REF2VA students; requires references and Mamad8's node package.\n\nThe Director's **Decoder** control chooses the original H3 Video VAE or the fastest experimental T=1 Image VAE. Temporal quality controls 5/9/13/20-frame video packets. Current ComfyUI chunked H3 VAE I/O reduces memory with identical pixels; it is not advertised as a speed boost."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 27,
"type": "H3StudioWorkflowNote",
"pos": [
820,
1660
],
"size": [
480,
400
],
"flags": {},
"order": 17,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Benchmark Lab - quality and speed diagnosis",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"optional / experimental",
"## Compare two or many setups\n\n1. Enter Base, LightX, or PDD profiles and resolution targets.\n2. Check the exact generation count shown on the node.\n3. Turn **Benchmark ON** in the purple Run mode switch.\n\nNormal sampling stays lazy and is not scheduled. Same seed is the fairest comparison; new seeds are for diversity sweeps. Disable live cells for maximum throughput."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 28,
"type": "H3StudioWorkflowNote",
"pos": [
1560,
1080
],
"size": [
430,
720
],
"flags": {},
"order": 18,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Model downloads · RECOMMENDED core",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"models",
"## RECOMMENDED · proven L4 core · ~50.4 GB / 47.0 GiB\n\nDownload each file directly into the shown `ComfyUI/models/` folder:\n\n- [Kijai FL2VA pruned W4A8 · 12.54 GB / 11.68 GiB](https://huggingface.co/Kijai/MiniMax-H3-experimental/resolve/main/minimax_h3_fl2va_pruned_w4a8_mixed.safetensors?download=true) → `diffusion_models/`\n- [Kijai REF2VA pruned W4A8 · 11.77 GB / 10.96 GiB](https://huggingface.co/Kijai/MiniMax-H3-experimental/resolve/main/minimax_h3_ref2va_pruned_w4a8_mixed.safetensors?download=true) → `diffusion_models/`\n- [H3 Qwen3-VL 32B NVFP4 · 15.7 GB](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors?download=true) → `text_encoders/`\n- [Original H3 Video VAE FP16 · 5.2 GB](https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors?download=true) → `vae/`\n- [Qwen3-VL 4B analyzer + writer · 5.2 GB](https://huggingface.co/Comfy-Org/Qwen3-VL/resolve/main/text_encoders/qwen3vl_4b_fp8_scaled.safetensors?download=true) → `text_encoders/`\n\n## Optional acceleration and previews\n\n- [TAEH3 live preview · 0.01 GB](https://huggingface.co/Kijai/MiniMax-H3-TAE/resolve/main/vae_approx/taeh3.safetensors?download=true) → `vae_approx/`\n- [LightX resized rank-21 · 0.31 GB](https://huggingface.co/Kijai/MiniMax-H3_comfy/resolve/main/loras/minimax_h3_fl2v_lightx2v_turbo_4step_v0.1_comfy_resized_avg_rank_21_bf16.safetensors?download=true) → `loras/`\n- [Experimental T=1 Image VAE · 5.2 GB](https://huggingface.co/Mamad8/MiniMax-H3-Image-VAE/resolve/main/minimax_h3_t1_image_vae_step1597.safetensors?download=true) → `vae/`\n- [Qwen3-VL 8B writer · 10.6 GB](https://huggingface.co/Comfy-Org/Qwen3-VL/resolve/main/text_encoders/qwen3vl_8b_fp8_scaled.safetensors?download=true) → `text_encoders/`\n\n## Optional PDD · choose 600 or 900 · ~1.30 GB each\n\n- [PDD 900 LoRA · 0.60 GB](https://huggingface.co/Mamad8/MiniMaxH3_R2V-PDD-Turbo-LoRA-Mamad8/resolve/main/LORA_h3_pdd_af384_step900_s.safetensors?download=true) → `loras/`\n- [PDD 900 heads · 0.70 GB](https://huggingface.co/Mamad8/MiniMaxH3_R2V-PDD-Turbo-LoRA-Mamad8/resolve/main/HEADS_h3_pdd_af384_step900_bank.safetensors?download=true) → `pdd_heads/`\n- [PDD 600 LoRA · 0.60 GB](https://huggingface.co/Mamad8/MiniMaxH3_R2V-PDD-Turbo-LoRA-Mamad8/resolve/main/LORA_h3_pdd_af384_step600_s.safetensors?download=true) → `loras/`\n- [PDD 600 heads · 0.70 GB](https://huggingface.co/Mamad8/MiniMaxH3_R2V-PDD-Turbo-LoRA-Mamad8/resolve/main/HEADS_h3_pdd_af384_step600_bank.safetensors?download=true) → `pdd_heads/`\n- [Install the required PDD custom node](https://github.com/mamad8c/ComfyUI-MiniMaxH3-PDD-Mamad8)\n\n> [!NOTE]\n> Core + every optional file above is ~69.2 GB. H3 Studio never downloads models automatically."
],
"color": "#22322f",
"bgcolor": "#17211f"
},
{
"id": 29,
"type": "H3StudioWorkflowNote",
"pos": [
1560,
1860
],
"size": [
480,
390
],
"flags": {},
"order": 19,
"mode": 0,
"inputs": [
{
"name": "section",
"type": "COMBO",
"link": null,
"widget": {
"name": "section"
}
},
{
"name": "text",
"type": "STRING",
"link": null,
"widget": {
"name": "text"
}
}
],
"outputs": [],
"title": "Upscaling, outpainting, and inpainting",
"properties": {
"Node name for S&R": "H3StudioWorkflowNote",
"purpose": "Documentation only; never enters the generation graph."
},
"widgets_values": [
"optional / experimental",
"Upscaling is safest after H3's selected still: use ComfyUI's built-in Upscale Model Loader + Image Upscale With Model and a trusted image upscaler. H3 semantic outpainting works by increasing the target aspect ratio and explicitly placing the original reference within the new frame. Exact mask inpainting is not implemented here: H3 Studio has no verified mask-conditioned H3 route, and ComfyUI's VAE Encode for Inpainting requires a model trained for that contract. The node will not pretend semantic reference editing is pixel-locked inpainting."
],
"color": "#22322f",
"bgcolor": "#17211f"
}
],
"links": [
[
1,
10,
0,
12,
1,
"H3_STUDIO_CONTEXT"
],
[
2,
11,
0,
12,
0,
"H3_STUDIO_BUNDLE"
],
[
3,
11,
0,
10,
23,
"H3_STUDIO_BUNDLE"
],
[
4,
12,
0,
16,
0,
"MODEL"
],
[
5,
16,
0,
13,
0,
"MODEL"
],
[
6,
12,
2,
13,
1,
"CONDITIONING"
],
[
7,
12,
3,
13,
2,
"LATENT"
],
[
8,
12,
4,
13,
3,
"VAE"
],
[
9,
10,
5,
13,
4,
"INT"
],
[
10,
10,
0,
13,
5,
"H3_STUDIO_CONTEXT"
],
[
11,
13,
0,
19,
1,
"IMAGE"
],
[
12,
11,
0,
17,
0,
"H3_STUDIO_BUNDLE"
],
[
13,
10,
0,
17,
1,
"H3_STUDIO_CONTEXT"
],
[
14,
10,
0,
15,
1,
"H3_STUDIO_CONTEXT"
],
[
15,
10,
0,
30,
1,
"H3_STUDIO_CONTEXT"
],
[
16,
17,
0,
19,
2,
"IMAGE"
],
[
17,
19,
0,
14,
0,
"IMAGE"
],
[
18,
19,
0,
15,
0,
"IMAGE"
],
[
19,
19,
0,
30,
0,
"IMAGE"
]
],
"groups": [
{
"id": 1,
"title": "01 · DIRECT THE IMAGE",
"bounding": [
-1500,
160,
860,
940
],
"color": "#355b54",
"font_size": 28,
"flags": {}
},
{
"id": 2,
"title": "02 · MODELS AND CONDITIONING",
"bounding": [
-580,
190,
680,
600
],
"color": "#2f665b",
"font_size": 26,
"flags": {}
},
{
"id": 3,
"title": "03 · SAMPLE AND EXTRACT STILL",
"bounding": [
160,
190,
740,
1230
],
"color": "#6b5937",
"font_size": 26,
"flags": {}
},
{
"id": 4,
"title": "04 · OUTPUT",
"bounding": [
970,
190,
560,
1320
],
"color": "#496073",
"font_size": 26,
"flags": {}
},
{
"id": 5,
"title": "REFERENCE GUIDE",
"bounding": [
1500,
190,
540,
2300
],
"color": "#5c5243",
"font_size": 24,
"flags": {}
},
{
"id": 6,
"title": "READ BEFORE FIRST RUN",
"bounding": [
-1500,
-240,
2750,
400
],
"color": "#324e49",
"font_size": 30,
"flags": {}
},
{
"id": 7,
"title": "OPTIONAL - MULTI-PROFILE BENCHMARK LAB",
"bounding": [
-580,
1600,
1940,
700
],
"color": "#60467a",
"font_size": 28,
"flags": {}
}
],
"definitions": {
"subgraphs": [
{
"id": "5930b00d-9f8e-4b87-9cb5-ff5f7cf3b30a",
"version": 1,
"state": {
"lastGroupId": 0,
"lastNodeId": 106,
"lastLinkId": 17,
"lastRerouteId": 0
},
"revision": 0,
"config": {},
"name": "H3 Studio · Sampling and Exact Still Decode",
"inputNode": {
"id": -10,
"bounding": [
-190,
40,
120,
360
]
},
"outputNode": {
"id": -20,
"bounding": [
2130,
120,
140,
240
]
},
"inputs": [
{
"id": "352b5c74-4db0-568b-9cb7-525414c5c991",
"name": "model",
"type": "MODEL",
"linkIds": [
1
],
"pos": [
-160,
132
]
},
{
"id": "337c26a9-6c85-5f64-9703-96e57ac16b8b",
"name": "positive",
"type": "CONDITIONING",
"linkIds": [
2
],
"pos": [
-160,
184
]
},
{
"id": "bcdb6c4e-a2cf-5492-89e0-5c021e653eb2",
"name": "h3_latent",
"type": "LATENT",
"linkIds": [
3
],
"pos": [
-160,
236
]
},
{
"id": "2ae2561c-b17e-5e42-810f-e4200b23d0aa",
"name": "video_vae",
"type": "VAE",
"linkIds": [
4
],
"pos": [
-160,
288
]
},
{
"id": "47028fc0-cd59-5d9f-bbdf-0284843c0dd8",
"name": "seed",
"type": "INT",
"linkIds": [
5
],
"pos": [
-160,
340
]
},
{
"id": "61cd865a-970b-507d-b878-275416e10d15",
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"linkIds": [
17
],
"pos": [
-160,
392
]
}
],
"outputs": [
{
"id": "fdf09fcf-6bbb-500e-ba22-ba94ef0f2587",
"name": "image",
"type": "IMAGE",
"linkIds": [
14
],
"pos": [
2160,
180
]
},
{
"id": "ff20728a-2f06-5d32-9b35-2a1a2934c2c3",
"name": "decode_info",
"type": "STRING",
"linkIds": [
15
],
"pos": [
2160,
232
]
},
{
"id": "6ad2c196-e415-5a30-9f8a-d8d29884da0a",
"name": "selection_report",
"type": "STRING",
"linkIds": [
16
],
"pos": [
2160,
284
]
}
],
"widgets": [],
"nodes": [
{
"id": 101,
"type": "H3StudioContextSamplingPreset",
"pos": [
40,
40
],
"size": [
340,
100
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 1
},
{
"name": "studio_context",
"type": "H3_STUDIO_CONTEXT",
"link": 17
}
],
"outputs": [
{
"name": "model",
"type": "MODEL",
"links": [
6
]
},
{
"name": "sampler",
"type": "SAMPLER",
"links": [
7
]
},
{
"name": "sigmas",
"type": "SIGMAS",
"links": [
8
]
},
{
"name": "sampling_info",
"type": "STRING",
"links": null
}
],
"title": "Director-selected sampling profile",
"properties": {
"Node name for S&R": "H3StudioContextSamplingPreset"
},
"widgets_values": []
},
{
"id": 102,
"type": "BasicGuider",
"pos": [
450,
60
],
"size": [
270,
72
],
"flags": {},
"order": 1,
"mode": 0,
"inputs": [
{
"name": "model",
"type": "MODEL",
"link": 6
},
{
"name": "conditioning",
"type": "CONDITIONING",
"link": 2
}
],
"outputs": [
{
"name": "GUIDER",
"type": "GUIDER",
"links": [
9
]
}
],
"title": "Positive guider",
"properties": {
"Node name for S&R": "BasicGuider"
},
"widgets_values": []
},
{
"id": 103,
"type": "RandomNoise",
"pos": [
50,
220
],
"size": [
280,
82
],
"flags": {},
"order": 2,
"mode": 0,
"inputs": [
{
"name": "noise_seed",
"type": "INT",
"link": 5,
"widget": {
"name": "noise_seed"
}
}
],
"outputs": [
{
"name": "NOISE",
"type": "NOISE",
"links": [
10
]
}
],
"title": "Studio seed",
"properties": {
"Node name for S&R": "RandomNoise"
},
"widgets_values": [
42,
"fixed"
]
},
{
"id": 104,
"type": "SamplerCustomAdvanced",
"pos": [
790,
150
],
"size": [
330,
150
],
"flags": {},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "noise",
"type": "NOISE",
"link": 10
},
{
"name": "guider",
"type": "GUIDER",
"link": 9
},
{
"name": "sampler",
"type": "SAMPLER",
"link": 7
},
{
"name": "sigmas",
"type": "SIGMAS",
"link": 8
},
{
"name": "latent_image",
"type": "LATENT",
"link": 3
}
],
"outputs": [
{
"name": "output",
"type": "LATENT",
"links": [
11
]
},
{
"name": "denoised_output",
"type": "LATENT",
"links": null
}
],
"title": "Sample H3 still packet",
"properties": {
"Node name for S&R": "SamplerCustomAdvanced"
},
"widgets_values": []
},
{
"id": 105,
"type": "H3StudioDecode",
"pos": [
1190,
150
],
"size": [
310,
104
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "samples",
"type": "LATENT",
"link": 11
},
{
"name": "vae",
"type": "VAE",
"link": 4
}
],
"outputs": [
{
"name": "frames",
"type": "IMAGE",
"links": [
12
]
},
{
"name": "decoded_frames",
"type": "INT",
"links": null
},
{
"name": "decode_info",
"type": "STRING",
"links": [
15
]
},
{
"name": "recommended_index",
"type": "INT",
"links": [
13
]
}
],
"title": "Exact frame decode",
"properties": {
"Node name for S&R": "H3StudioDecode"
},
"widgets_values": []
},
{
"id": 106,
"type": "H3StudioFrameSelector",
"pos": [
1570,
100
],
"size": [
430,
360
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "frames",
"type": "IMAGE",
"link": 12
},
{
"name": "strategy",
"type": "COMBO",
"link": null,
"widget": {
"name": "strategy"
}
},
{
"name": "manual_index",
"type": "INT",
"link": null,
"widget": {
"name": "manual_index"
}
},
{
"name": "skip_first_frames",
"type": "INT",
"link": null,
"widget": {
"name": "skip_first_frames"
}
},
{
"name": "candidate_start",
"type": "FLOAT",
"link": null,
"widget": {
"name": "candidate_start"
}
},
{
"name": "candidate_end",
"type": "FLOAT",
"link": null,
"widget": {
"name": "candidate_end"
}
},
{
"name": "similarity_weight",
"type": "FLOAT",
"link": null,
"widget": {
"name": "similarity_weight"
}
},
{
"name": "top_k",
"type": "INT",
"link": null,
"widget": {
"name": "top_k"
}
},
{
"name": "source_image",
"type": "IMAGE",
"link": null
},
{
"name": "emit_candidate_batch",
"type": "BOOLEAN",
"link": null,
"widget": {
"name": "emit_candidate_batch"
}
},
{
"name": "recommended_index",
"type": "INT",
"link": 13
}
],
"outputs": [
{
"name": "selected_image",
"type": "IMAGE",
"links": [
14
]
},
{
"name": "candidate_batch_debug",
"type": "IMAGE",
"links": null
},
{
"name": "selected_index",
"type": "INT",
"links": null
},
{
"name": "selected_score",
"type": "FLOAT",
"links": null
},
{
"name": "selection_report",
"type": "STRING",
"links": [
16
]
}
],
"title": "Select final still",
"properties": {
"Node name for S&R": "H3StudioFrameSelector"
},
"widgets_values": [
"decode_recommended",
0,
0,
0.0,
1.0,
0.6,
4,
false
]
}
],
"groups": [
{
"id": 1,
"title": "01 · H3 sampling",
"bounding": [
10,
0,
740,
350
],
"color": "#355b54",
"font_size": 22,
"flags": {}
},
{
"id": 2,
"title": "02 · Decode the temporal packet",
"bounding": [
760,
20,
760,
330
],
"color": "#3f536b",
"font_size": 22,
"flags": {}
},
{
"id": 3,
"title": "03 · Choose one still",
"bounding": [
1530,
20,
520,
490
],
"color": "#66553d",
"font_size": 22,
"flags": {}
}
],
"links": [
{
"id": 1,
"origin_id": -10,
"origin_slot": 0,
"target_id": 101,
"target_slot": 0,
"type": "MODEL"
},
{
"id": 2,
"origin_id": -10,
"origin_slot": 1,
"target_id": 102,
"target_slot": 1,
"type": "CONDITIONING"
},
{
"id": 3,
"origin_id": -10,
"origin_slot": 2,
"target_id": 104,
"target_slot": 4,
"type": "LATENT"
},
{
"id": 4,
"origin_id": -10,
"origin_slot": 3,
"target_id": 105,
"target_slot": 1,
"type": "VAE"
},
{
"id": 5,
"origin_id": -10,
"origin_slot": 4,
"target_id": 103,
"target_slot": 0,
"type": "INT"
},
{
"id": 6,
"origin_id": 101,
"origin_slot": 0,
"target_id": 102,
"target_slot": 0,
"type": "MODEL"
},
{
"id": 7,
"origin_id": 101,
"origin_slot": 1,
"target_id": 104,
"target_slot": 2,
"type": "SAMPLER"
},
{
"id": 8,
"origin_id": 101,
"origin_slot": 2,
"target_id": 104,
"target_slot": 3,
"type": "SIGMAS"
},
{
"id": 9,
"origin_id": 102,
"origin_slot": 0,
"target_id": 104,
"target_slot": 1,
"type": "GUIDER"
},
{
"id": 10,
"origin_id": 103,
"origin_slot": 0,
"target_id": 104,
"target_slot": 0,
"type": "NOISE"
},
{
"id": 11,
"origin_id": 104,
"origin_slot": 0,
"target_id": 105,
"target_slot": 0,
"type": "LATENT"
},
{
"id": 12,
"origin_id": 105,
"origin_slot": 0,
"target_id": 106,
"target_slot": 0,
"type": "IMAGE"
},
{
"id": 13,
"origin_id": 105,
"origin_slot": 3,
"target_id": 106,
"target_slot": 10,
"type": "INT"
},
{
"id": 14,
"origin_id": 106,
"origin_slot": 0,
"target_id": -20,
"target_slot": 0,
"type": "IMAGE"
},
{
"id": 15,
"origin_id": 105,
"origin_slot": 2,
"target_id": -20,
"target_slot": 1,
"type": "STRING"
},
{
"id": 16,
"origin_id": 106,
"origin_slot": 4,
"target_id": -20,
"target_slot": 2,
"type": "STRING"
},
{
"id": 17,
"origin_id": -10,
"origin_slot": 5,
"target_id": 101,
"target_slot": 1,
"type": "H3_STUDIO_CONTEXT"
}
],
"extra": {
"documentation": "The Director stays outside this subgraph because ComfyUI does not reliably promote custom DOM widgets."
},
"category": "H3 Studio",
"description": "Director-selected H3 sampling, exact temporal decode, and stable still selection."
}
]
},
"config": {},
"extra": {
"ds": {
"scale": 0.72,
"offset": [
1120,
330
]
},
"frontendVersion": "1.30.0",
"h3studio": {
"schema_version": 8,
"template_version": "1.6.0",
"design_source": "H3 Studio maintained column layout",
"audio_prompt_sections": false,
"hub_included": false
}
},
"version": 0.4
}