Pika H3 Flex: Fast test opens plainly off, and examples can't ship a half-set switch

The Fast test row set the video loader to 1 MP when off, but the workflow ships the loader at 0.41 MP, so the row opened striped 'changed' with neither side picked. Its off side is now 0.41.

workflow_contract now fails any Workflow Switches row whose saved values sit on neither side, are half on and half off, or disagree with whether its nodes run.

example_workflows/Pika H3 Flex (AusBoss).json

@@ -39,10 +39,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "{\"v\": 1, \"title\": \"Pika H3 Flex Video | Swap a Person, Copy Moves, Voice and Song\", \"subtitle\": \"Put your person into a video, or type what happens and give each picture a job. Add a voice or a song when you want one.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"1. Load **Picture 1**.\\n2. **Type what happens**, one line for each thing.\\n3. Pick how many **seconds** on **Seconds**.\\n4. Press **Run**.\\n\\nThat is all you need. Everything else is a switch in **Pictures**, **Video and sound** or **Extras**, and it stays out of the way until you turn it on.\\n\\n**The first run takes longer.** H3, its text encoder and its VAEs come to about 45 GB, and all of it has to load once.\\n\\nEach part of the page has its own note with the details: **Settings**, **pictures**, **recipes**, **video and sound**, and **writing the lines**.\\n\\nYou never have to learn H3's format. H3 Prompt Writer writes it for you, and **Prompt H3 sees** on the right shows what it wrote.\\n\\nPika H3 Flex is by Pika, on MiniMax H3 by MiniMax. These are Comfy-Org's ComfyUI files.\\n\\nNeeds ComfyUI {COMFY_NEEDS} or newer and AusBoss nodes {PACK} or newer (update first).\", \"models\": [{\"name\": \"minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"20.97 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"note\": \"the Ref2VA model, not FL2VA\"}, {\"name\": \"qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"dir\": \"text_encoders\", \"size\": \"15.69 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"note\": \"the same file every H3 workflow uses: you may have it already\"}, {\"name\": \"minimax_h3_video_vae_fp16.safetensors\", \"dir\": \"vae\", \"size\": \"5.21 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors\", \"note\": \"\"}, {\"name\": \"minimax_h3_audio_vae_fp32.safetensors\", \"dir\": \"vae\", \"size\": \"0.60 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors\", \"note\": \"the sound\"}, {\"name\": \"pika_h3_flex_ref2va_v1_bf16.safetensors\", \"dir\": \"loras\", \"size\": \"1.96 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/pika_h3_flex_ref2va_v1_bf16.safetensors\", \"note\": \"Pika's add-on, already in the LoRA Loader\"}, {\"name\": \"minimax_h3_ref2va_pruned_w6a8.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"15.98 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_w6a8.safetensors\", \"note\": \"only for Low VRAM: use it in place of the first file\"}, {\"name\": \"minimax_h3_video_vae_int8_convrot.safetensors\", \"dir\": \"vae\", \"size\": \"2.81 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors\", \"note\": \"only for Low VRAM: use it in place of the VAE above\"}, {\"name\": \"qwen3vl_8b_int8_convrot.safetensors\", \"dir\": \"text_encoders\", \"size\": \"9.35 GB\", \"url\": \"https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors\", \"note\": \"only for Auto prompt: the small text model\"}], \"packs\": [{\"name\": \"ComfyUI-AusBoss\", \"url\": \"https://github.com/ausboss/ComfyUI-AusBoss\", \"node\": \"AUSBOSS_NODES_Callout\"}], \"links\": [{\"label\": \"YouTube \u00b7 20 things you can do with an edit model\", \"url\": \"https://youtu.be/GociGWQQDSs\"}, {\"label\": \"YouTube \u00b7 ausbossdev\", \"url\": \"https://www.youtube.com/@ausbossdev\"}, {\"label\": \"Pika H3 Flex \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Pika-Labs/Pika-H3-Flex\"}, {\"label\": \"MiniMax H3 files \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3\"}, {\"label\": \"GitHub \u00b7 ausboss\", \"url\": \"https://github.com/ausboss\"}, {\"label\": \"Civitai \u00b7 AusBoss\", \"url\": \"https://civitai.com/user/AusBoss\"}, {\"label\": \"X \u00b7 Zanzibased\", \"url\": \"https://x.com/Zanzibased\"}]}"
+    "{\"v\": 1, \"title\": \"Pika H3 Flex Video | Swap a Person, Copy Moves, Voice and Song\", \"subtitle\": \"Put your person into a video, or type what happens and give each picture a job. Add a voice or a song when you want one.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"1. Load **Picture 1**.\\n2. **Type what happens**, one line for each thing.\\n3. Pick how many **seconds** on **Seconds**.\\n4. Press **Run**.\\n\\nThat is all you need. Everything else is a switch in **Pictures**, **Video and sound** or **Extras**, and it stays out of the way until you turn it on.\\n\\n**The first run takes longer.** H3, its text encoder and its VAEs come to about 45 GB, and all of it has to load once.\\n\\nEach part of the page has its own note with the details: **Settings**, **pictures**, **recipes**, **video and sound**, and **writing the lines**.\\n\\nYou never have to learn H3's format. H3 Prompt Writer writes it for you, and **Prompt H3 sees** on the right shows what it wrote.\\n\\nPika H3 Flex is by Pika, on MiniMax H3 by MiniMax. These are Comfy-Org's ComfyUI files.\", \"models\": [{\"name\": \"minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"20.97 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"note\": \"the Ref2VA model, not FL2VA\"}, {\"name\": \"qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"dir\": \"text_encoders\", \"size\": \"15.69 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"note\": \"the same file every H3 workflow uses: you may have it already\"}, {\"name\": \"minimax_h3_video_vae_fp16.safetensors\", \"dir\": \"vae\", \"size\": \"5.21 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors\", \"note\": \"\"}, {\"name\": \"minimax_h3_audio_vae_fp32.safetensors\", \"dir\": \"vae\", \"size\": \"0.60 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors\", \"note\": \"the sound\"}, {\"name\": \"pika_h3_flex_ref2va_v1_bf16.safetensors\", \"dir\": \"loras\", \"size\": \"1.96 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/pika_h3_flex_ref2va_v1_bf16.safetensors\", \"note\": \"Pika's add-on, already in the LoRA Loader\"}, {\"name\": \"minimax_h3_ref2va_pruned_w6a8.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"15.98 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_w6a8.safetensors\", \"note\": \"only for Low VRAM: use it in place of the first file\"}, {\"name\": \"minimax_h3_video_vae_int8_convrot.safetensors\", \"dir\": \"vae\", \"size\": \"2.81 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors\", \"note\": \"only for Low VRAM: use it in place of the VAE above\"}, {\"name\": \"qwen3vl_8b_int8_convrot.safetensors\", \"dir\": \"text_encoders\", \"size\": \"9.35 GB\", \"url\": \"https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors\", \"note\": \"only for Auto prompt: the small text model\"}], \"packs\": [{\"name\": \"ComfyUI-AusBoss\", \"url\": \"https://github.com/ausboss/ComfyUI-AusBoss\", \"node\": \"AUSBOSS_NODES_Callout\"}], \"links\": [{\"label\": \"YouTube \u00b7 20 things you can do with an edit model\", \"url\": \"https://youtu.be/GociGWQQDSs\"}, {\"label\": \"YouTube \u00b7 ausbossdev\", \"url\": \"https://www.youtube.com/@ausbossdev\"}, {\"label\": \"Pika H3 Flex \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Pika-Labs/Pika-H3-Flex\"}, {\"label\": \"MiniMax H3 files \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3\"}, {\"label\": \"GitHub \u00b7 ausboss\", \"url\": \"https://github.com/ausboss\"}, {\"label\": \"Civitai \u00b7 AusBoss\", \"url\": \"https://civitai.com/user/AusBoss\"}, {\"label\": \"X \u00b7 Zanzibased\", \"url\": \"https://x.com/Zanzibased\"}]}"
    ],
    "widgets_values_named": {
-    "note": "{\"v\": 1, \"title\": \"Pika H3 Flex Video | Swap a Person, Copy Moves, Voice and Song\", \"subtitle\": \"Put your person into a video, or type what happens and give each picture a job. Add a voice or a song when you want one.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"1. Load **Picture 1**.\\n2. **Type what happens**, one line for each thing.\\n3. Pick how many **seconds** on **Seconds**.\\n4. Press **Run**.\\n\\nThat is all you need. Everything else is a switch in **Pictures**, **Video and sound** or **Extras**, and it stays out of the way until you turn it on.\\n\\n**The first run takes longer.** H3, its text encoder and its VAEs come to about 45 GB, and all of it has to load once.\\n\\nEach part of the page has its own note with the details: **Settings**, **pictures**, **recipes**, **video and sound**, and **writing the lines**.\\n\\nYou never have to learn H3's format. H3 Prompt Writer writes it for you, and **Prompt H3 sees** on the right shows what it wrote.\\n\\nPika H3 Flex is by Pika, on MiniMax H3 by MiniMax. These are Comfy-Org's ComfyUI files.\\n\\nNeeds ComfyUI {COMFY_NEEDS} or newer and AusBoss nodes {PACK} or newer (update first).\", \"models\": [{\"name\": \"minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"20.97 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"note\": \"the Ref2VA model, not FL2VA\"}, {\"name\": \"qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"dir\": \"text_encoders\", \"size\": \"15.69 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"note\": \"the same file every H3 workflow uses: you may have it already\"}, {\"name\": \"minimax_h3_video_vae_fp16.safetensors\", \"dir\": \"vae\", \"size\": \"5.21 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors\", \"note\": \"\"}, {\"name\": \"minimax_h3_audio_vae_fp32.safetensors\", \"dir\": \"vae\", \"size\": \"0.60 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors\", \"note\": \"the sound\"}, {\"name\": \"pika_h3_flex_ref2va_v1_bf16.safetensors\", \"dir\": \"loras\", \"size\": \"1.96 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/pika_h3_flex_ref2va_v1_bf16.safetensors\", \"note\": \"Pika's add-on, already in the LoRA Loader\"}, {\"name\": \"minimax_h3_ref2va_pruned_w6a8.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"15.98 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_w6a8.safetensors\", \"note\": \"only for Low VRAM: use it in place of the first file\"}, {\"name\": \"minimax_h3_video_vae_int8_convrot.safetensors\", \"dir\": \"vae\", \"size\": \"2.81 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors\", \"note\": \"only for Low VRAM: use it in place of the VAE above\"}, {\"name\": \"qwen3vl_8b_int8_convrot.safetensors\", \"dir\": \"text_encoders\", \"size\": \"9.35 GB\", \"url\": \"https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors\", \"note\": \"only for Auto prompt: the small text model\"}], \"packs\": [{\"name\": \"ComfyUI-AusBoss\", \"url\": \"https://github.com/ausboss/ComfyUI-AusBoss\", \"node\": \"AUSBOSS_NODES_Callout\"}], \"links\": [{\"label\": \"YouTube \u00b7 20 things you can do with an edit model\", \"url\": \"https://youtu.be/GociGWQQDSs\"}, {\"label\": \"YouTube \u00b7 ausbossdev\", \"url\": \"https://www.youtube.com/@ausbossdev\"}, {\"label\": \"Pika H3 Flex \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Pika-Labs/Pika-H3-Flex\"}, {\"label\": \"MiniMax H3 files \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3\"}, {\"label\": \"GitHub \u00b7 ausboss\", \"url\": \"https://github.com/ausboss\"}, {\"label\": \"Civitai \u00b7 AusBoss\", \"url\": \"https://civitai.com/user/AusBoss\"}, {\"label\": \"X \u00b7 Zanzibased\", \"url\": \"https://x.com/Zanzibased\"}]}"
+    "note": "{\"v\": 1, \"title\": \"Pika H3 Flex Video | Swap a Person, Copy Moves, Voice and Song\", \"subtitle\": \"Put your person into a video, or type what happens and give each picture a job. Add a voice or a song when you want one.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"1. Load **Picture 1**.\\n2. **Type what happens**, one line for each thing.\\n3. Pick how many **seconds** on **Seconds**.\\n4. Press **Run**.\\n\\nThat is all you need. Everything else is a switch in **Pictures**, **Video and sound** or **Extras**, and it stays out of the way until you turn it on.\\n\\n**The first run takes longer.** H3, its text encoder and its VAEs come to about 45 GB, and all of it has to load once.\\n\\nEach part of the page has its own note with the details: **Settings**, **pictures**, **recipes**, **video and sound**, and **writing the lines**.\\n\\nYou never have to learn H3's format. H3 Prompt Writer writes it for you, and **Prompt H3 sees** on the right shows what it wrote.\\n\\nPika H3 Flex is by Pika, on MiniMax H3 by MiniMax. These are Comfy-Org's ComfyUI files.\", \"models\": [{\"name\": \"minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"20.97 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_int8_convrot.safetensors\", \"note\": \"the Ref2VA model, not FL2VA\"}, {\"name\": \"qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"dir\": \"text_encoders\", \"size\": \"15.69 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/text_encoders/qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors\", \"note\": \"the same file every H3 workflow uses: you may have it already\"}, {\"name\": \"minimax_h3_video_vae_fp16.safetensors\", \"dir\": \"vae\", \"size\": \"5.21 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_fp16.safetensors\", \"note\": \"\"}, {\"name\": \"minimax_h3_audio_vae_fp32.safetensors\", \"dir\": \"vae\", \"size\": \"0.60 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_audio_vae_fp32.safetensors\", \"note\": \"the sound\"}, {\"name\": \"pika_h3_flex_ref2va_v1_bf16.safetensors\", \"dir\": \"loras\", \"size\": \"1.96 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/loras/pika_h3_flex_ref2va_v1_bf16.safetensors\", \"note\": \"Pika's add-on, already in the LoRA Loader\"}, {\"name\": \"minimax_h3_ref2va_pruned_w6a8.safetensors\", \"dir\": \"diffusion_models\", \"size\": \"15.98 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/diffusion_models/minimax_h3_ref2va_pruned_w6a8.safetensors\", \"note\": \"only for Low VRAM: use it in place of the first file\"}, {\"name\": \"minimax_h3_video_vae_int8_convrot.safetensors\", \"dir\": \"vae\", \"size\": \"2.81 GB\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3/resolve/main/vae/minimax_h3_video_vae_int8_convrot.safetensors\", \"note\": \"only for Low VRAM: use it in place of the VAE above\"}, {\"name\": \"qwen3vl_8b_int8_convrot.safetensors\", \"dir\": \"text_encoders\", \"size\": \"9.35 GB\", \"url\": \"https://huggingface.co/Comfy-Org/Qwen-Image-2.1/resolve/main/text_encoders/qwen3vl_8b_int8_convrot.safetensors\", \"note\": \"only for Auto prompt: the small text model\"}], \"packs\": [{\"name\": \"ComfyUI-AusBoss\", \"url\": \"https://github.com/ausboss/ComfyUI-AusBoss\", \"node\": \"AUSBOSS_NODES_Callout\"}], \"links\": [{\"label\": \"YouTube \u00b7 20 things you can do with an edit model\", \"url\": \"https://youtu.be/GociGWQQDSs\"}, {\"label\": \"YouTube \u00b7 ausbossdev\", \"url\": \"https://www.youtube.com/@ausbossdev\"}, {\"label\": \"Pika H3 Flex \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Pika-Labs/Pika-H3-Flex\"}, {\"label\": \"MiniMax H3 files \u00b7 Hugging Face\", \"url\": \"https://huggingface.co/Comfy-Org/MiniMax-H3\"}, {\"label\": \"GitHub \u00b7 ausboss\", \"url\": \"https://github.com/ausboss\"}, {\"label\": \"Civitai \u00b7 AusBoss\", \"url\": \"https://civitai.com/user/AusBoss\"}, {\"label\": \"X \u00b7 Zanzibased\", \"url\": \"https://x.com/Zanzibased\"}]}"
    },
    "color": "#14424d",
    "bgcolor": "#161f21"
@@ -1456,14 +1456,14 @@
     "ver": "0.43.2"
    },
    "widgets_values": [
-    "01_card_declined.mp4",
+    "",
     0,
     0,
     10,
     ""
    ],
    "widgets_values_named": {
-    "audio": "01_card_declined.mp4",
+    "audio": "",
     "start_seconds": 0,
     "end_seconds": 0,
     "max_seconds": 10,
@@ -1566,14 +1566,14 @@
     "ver": "0.43.2"
    },
    "widgets_values": [
-    "01_card_declined.mp4",
+    "",
     0,
     0,
     10,
     ""
    ],
    "widgets_values_named": {
-    "audio": "01_card_declined.mp4",
+    "audio": "",
     "start_seconds": 0,
     "end_seconds": 0,
     "max_seconds": 10,
@@ -1983,7 +1983,7 @@
      "links": []
     }
    ],
-   "title": "Video - Swap or copy moves",
+   "title": "Video - Your clip",
    "properties": {
     "Node name for S&R": "AUSBOSS_NODES_VideoCropRotatePadClip",
     "cnr_id": "ausboss-nodes",
@@ -3452,10 +3452,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "{\"v\": 1, \"title\": \"Settings\", \"subtitle\": \"Only Picture 1 is on as it opens. Load it, type, press Run.\", \"author\": \"\", \"body\": \"### Swap someone in a video\\n1. Switch on **Video** and **Keep the clip's sound**, and load your clip.\\n2. Load the new person in **Picture 1**.\\n3. Clear **What happens** and press **Run**.\\n\\nTwo people in the clip? Type who goes, like \\\"the man in the hat\\\".\\n\\n### Good to know\\n- Click an arrow on a note to jump to what it points at.\\n- Purple nodes are switched off and need no file. The switches turn them on.\\n\\n### A clip came out wrong?\\nRun it again: the seed changes after every run. Still wrong? Read **Prompt H3 sees** on the right, then say it more plainly, one thing at a time. \\\"Value not in list\\\" means a file is missing.\", \"models\": [], \"packs\": [], \"links\": [], \"layout\": \"card\", \"accent\": \"\"}"
+    "{\"v\": 1, \"title\": \"Settings\", \"subtitle\": \"Only Picture 1 is on as it opens. Pick what you are making, then press Run.\", \"author\": \"\", \"body\": \"### Make a clip from pictures\\n1. Load **Picture 1** and pick its job on **H3 Prompt Writer**: person, place, outfit or thing.\\n2. Type what happens, one line for each thing.\\n3. Press **Run**.\\n\\n### Swap someone in a video\\n1. Switch on **Video** and **Keep the clip's sound**, and load your clip.\\n2. Load the new person in **Picture 1**.\\n3. Set the video job to **swap the person** on **H3 Prompt Writer**.\\n4. Clear **What happens** and press **Run**.\\n\\nSwapping two people? Switch on **Picture 2** too: Picture 1 takes the left one's place and Picture 2 the right one's. Or type who goes, one line each: `@1 the man in the hat`. Two can take a couple of runs.\\n\\n### Put a hat or an outfit on the person in your video\\n1. Switch on **Video**, **Keep the clip's sound** and **Picture 2**, and load your clip.\\n2. Load a still of the person from the video in **Picture 1**. A pasted screenshot works. Load the hat or outfit in **Picture 2**.\\n3. On **H3 Prompt Writer**, pick person for Picture 1, outfit for Picture 2, and the video job **swap the person**.\\n4. Type `@1 wears @2 in place of his cap` and press **Run**.\\n\\n### Good to know\\n- Click an arrow on a note to jump to what it points at.\\n- Purple nodes are switched off and need no file. The switches turn them on.\\n\\n### A clip came out wrong?\\nRun it again: the seed changes after every run. Still wrong? Read **Prompt H3 sees** on the right, then say it more plainly, one thing at a time. \\\"Value not in list\\\" means a file is missing.\", \"models\": [], \"packs\": [], \"links\": [], \"layout\": \"card\", \"accent\": \"\"}"
    ],
    "widgets_values_named": {
-    "note": "{\"v\": 1, \"title\": \"Settings\", \"subtitle\": \"Only Picture 1 is on as it opens. Load it, type, press Run.\", \"author\": \"\", \"body\": \"### Swap someone in a video\\n1. Switch on **Video** and **Keep the clip's sound**, and load your clip.\\n2. Load the new person in **Picture 1**.\\n3. Clear **What happens** and press **Run**.\\n\\nTwo people in the clip? Type who goes, like \\\"the man in the hat\\\".\\n\\n### Good to know\\n- Click an arrow on a note to jump to what it points at.\\n- Purple nodes are switched off and need no file. The switches turn them on.\\n\\n### A clip came out wrong?\\nRun it again: the seed changes after every run. Still wrong? Read **Prompt H3 sees** on the right, then say it more plainly, one thing at a time. \\\"Value not in list\\\" means a file is missing.\", \"models\": [], \"packs\": [], \"links\": [], \"layout\": \"card\", \"accent\": \"\"}"
+    "note": "{\"v\": 1, \"title\": \"Settings\", \"subtitle\": \"Only Picture 1 is on as it opens. Pick what you are making, then press Run.\", \"author\": \"\", \"body\": \"### Make a clip from pictures\\n1. Load **Picture 1** and pick its job on **H3 Prompt Writer**: person, place, outfit or thing.\\n2. Type what happens, one line for each thing.\\n3. Press **Run**.\\n\\n### Swap someone in a video\\n1. Switch on **Video** and **Keep the clip's sound**, and load your clip.\\n2. Load the new person in **Picture 1**.\\n3. Set the video job to **swap the person** on **H3 Prompt Writer**.\\n4. Clear **What happens** and press **Run**.\\n\\nSwapping two people? Switch on **Picture 2** too: Picture 1 takes the left one's place and Picture 2 the right one's. Or type who goes, one line each: `@1 the man in the hat`. Two can take a couple of runs.\\n\\n### Put a hat or an outfit on the person in your video\\n1. Switch on **Video**, **Keep the clip's sound** and **Picture 2**, and load your clip.\\n2. Load a still of the person from the video in **Picture 1**. A pasted screenshot works. Load the hat or outfit in **Picture 2**.\\n3. On **H3 Prompt Writer**, pick person for Picture 1, outfit for Picture 2, and the video job **swap the person**.\\n4. Type `@1 wears @2 in place of his cap` and press **Run**.\\n\\n### Good to know\\n- Click an arrow on a note to jump to what it points at.\\n- Purple nodes are switched off and need no file. The switches turn them on.\\n\\n### A clip came out wrong?\\nRun it again: the seed changes after every run. Still wrong? Read **Prompt H3 sees** on the right, then say it more plainly, one thing at a time. \\\"Value not in list\\\" means a file is missing.\", \"models\": [], \"packs\": [], \"links\": [], \"layout\": \"card\", \"accent\": \"\"}"
    },
    "color": "#14424d",
    "bgcolor": "#161f21"
@@ -3495,10 +3495,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "{\"v\": 1, \"title\": \"Your pictures\", \"subtitle\": \"Give each picture one job on H3 Prompt Writer. Type @1 for Picture 1, @2 for Picture 2.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **person**: a clear photo of one person. They go wherever your lines say. Name no place and they stay where the picture shows.\\n- **character sheet**: for a swap, crop it to one view with the blue handles. A whole sheet can flash up at the start, and a second picture of the same person becomes a second person.\\n- **place**: a place with nobody in it. Pale bars at the sides of the clip? Crop this picture to the clip's shape with the blue handles.\\n- **outfit** or **thing**: nobody in the picture. A person in it pulls the face toward theirs, so crop the head off with the blue handles.\\n- **style**: only the look is used.\\n- **start on it** or **end on it**: the first or the last moment of the clip.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "{\"v\": 1, \"title\": \"Your pictures\", \"subtitle\": \"Give each picture one job on H3 Prompt Writer, under the words. Picture 1 opens as person, so change it when you load a place, a hat or a thing. Type @1 for Picture 1, @2 for Picture 2.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **person**: a clear photo of one person. They go wherever your lines say. Name no place and they stay where the picture shows.\\n- **character sheet**: for a swap, crop it to one view with the blue handles. A whole sheet can flash up at the start, and a second picture of the same person becomes a second person.\\n- **place**: a place with nobody in it. Pale bars at the sides of the clip? Crop this picture to the clip's shape with the blue handles.\\n- **outfit** or **thing**: nobody in the picture. A person in it pulls the face toward theirs, so crop the head off with the blue handles. To put it on the person in your video, see **A new hat or outfit** in Recipes.\\n- **style**: only the look is used.\\n- **start on it** or **end on it**: the first or the last moment of the clip.\", \"models\": [], \"packs\": [], \"links\": []}"
    ],
    "widgets_values_named": {
-    "note": "{\"v\": 1, \"title\": \"Your pictures\", \"subtitle\": \"Give each picture one job on H3 Prompt Writer. Type @1 for Picture 1, @2 for Picture 2.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **person**: a clear photo of one person. They go wherever your lines say. Name no place and they stay where the picture shows.\\n- **character sheet**: for a swap, crop it to one view with the blue handles. A whole sheet can flash up at the start, and a second picture of the same person becomes a second person.\\n- **place**: a place with nobody in it. Pale bars at the sides of the clip? Crop this picture to the clip's shape with the blue handles.\\n- **outfit** or **thing**: nobody in the picture. A person in it pulls the face toward theirs, so crop the head off with the blue handles.\\n- **style**: only the look is used.\\n- **start on it** or **end on it**: the first or the last moment of the clip.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "note": "{\"v\": 1, \"title\": \"Your pictures\", \"subtitle\": \"Give each picture one job on H3 Prompt Writer, under the words. Picture 1 opens as person, so change it when you load a place, a hat or a thing. Type @1 for Picture 1, @2 for Picture 2.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **person**: a clear photo of one person. They go wherever your lines say. Name no place and they stay where the picture shows.\\n- **character sheet**: for a swap, crop it to one view with the blue handles. A whole sheet can flash up at the start, and a second picture of the same person becomes a second person.\\n- **place**: a place with nobody in it. Pale bars at the sides of the clip? Crop this picture to the clip's shape with the blue handles.\\n- **outfit** or **thing**: nobody in the picture. A person in it pulls the face toward theirs, so crop the head off with the blue handles. To put it on the person in your video, see **A new hat or outfit** in Recipes.\\n- **style**: only the look is used.\\n- **start on it** or **end on it**: the first or the last moment of the clip.\", \"models\": [], \"packs\": [], \"links\": []}"
    },
    "color": "#14424d",
    "bgcolor": "#161f21"
@@ -3538,10 +3538,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "{\"v\": 1, \"title\": \"Recipes\", \"subtitle\": \"What to turn on, and what to type.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **A story across places.** Nothing extra. One line for each place, with a blank line between.\\n- **Your own place.** Picture 2 as place: `@1 sits at the counter in @2`\\n- **Two people.** Picture 2 as person: `@1 and @2 sit side by side on a night bus`\\n- **A new outfit.** Picture 2 as outfit: `@1 walks down a runway wearing @2`\\n- **A product film.** Picture 1 as thing: `@1 spins slowly on black glass`\\n- **Copy a dance.** Video on, as copy its moves: `@1 dances on a rooftop at sunset`\\n- **That dance, that outfit.** Video on, Picture 2 as outfit: `@1 dances on a rooftop wearing @2`\\n- **Put someone in a video.** Video on, as swap the person. Type nothing.\\n- **A longer story.** Video on, as keep going. Load your last clip and type what comes next.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "{\"v\": 1, \"title\": \"Recipes\", \"subtitle\": \"What to turn on, and what to type.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **A story across places.** Nothing extra. One line for each place, with a blank line between.\\n- **Your own place.** Picture 2 as place: `@1 sits at the counter in @2`\\n- **Two people.** Picture 2 as person: `@1 and @2 sit side by side on a night bus`\\n- **A new outfit.** Picture 2 as outfit: `@1 walks down a runway wearing @2`\\n- **A product film.** Picture 1 as thing: `@1 spins slowly on black glass`\\n- **Copy a dance.** Video on, as copy its moves. Picture 1 is your person: `@1 dances on a rooftop at sunset`\\n- **That dance, that outfit.** Video on, as copy its moves. Picture 1 is your person, Picture 2 as outfit: `@1 dances on a rooftop wearing @2`\\n- **Put someone in a video.** Video on, as swap the person. Picture 1 is the new person. Type nothing.\\n- **A new hat or outfit on the person in your video.** Video on, as swap the person. Picture 1 is a still of them from the video (a pasted screenshot works), as person. Picture 2 is the hat or outfit, as outfit: `@1 wears @2 in place of his cap`\\n- **A longer story.** Video on, as keep going. Load your last clip and type what comes next.\", \"models\": [], \"packs\": [], \"links\": []}"
    ],
    "widgets_values_named": {
-    "note": "{\"v\": 1, \"title\": \"Recipes\", \"subtitle\": \"What to turn on, and what to type.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **A story across places.** Nothing extra. One line for each place, with a blank line between.\\n- **Your own place.** Picture 2 as place: `@1 sits at the counter in @2`\\n- **Two people.** Picture 2 as person: `@1 and @2 sit side by side on a night bus`\\n- **A new outfit.** Picture 2 as outfit: `@1 walks down a runway wearing @2`\\n- **A product film.** Picture 1 as thing: `@1 spins slowly on black glass`\\n- **Copy a dance.** Video on, as copy its moves: `@1 dances on a rooftop at sunset`\\n- **That dance, that outfit.** Video on, Picture 2 as outfit: `@1 dances on a rooftop wearing @2`\\n- **Put someone in a video.** Video on, as swap the person. Type nothing.\\n- **A longer story.** Video on, as keep going. Load your last clip and type what comes next.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "note": "{\"v\": 1, \"title\": \"Recipes\", \"subtitle\": \"What to turn on, and what to type.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"- **A story across places.** Nothing extra. One line for each place, with a blank line between.\\n- **Your own place.** Picture 2 as place: `@1 sits at the counter in @2`\\n- **Two people.** Picture 2 as person: `@1 and @2 sit side by side on a night bus`\\n- **A new outfit.** Picture 2 as outfit: `@1 walks down a runway wearing @2`\\n- **A product film.** Picture 1 as thing: `@1 spins slowly on black glass`\\n- **Copy a dance.** Video on, as copy its moves. Picture 1 is your person: `@1 dances on a rooftop at sunset`\\n- **That dance, that outfit.** Video on, as copy its moves. Picture 1 is your person, Picture 2 as outfit: `@1 dances on a rooftop wearing @2`\\n- **Put someone in a video.** Video on, as swap the person. Picture 1 is the new person. Type nothing.\\n- **A new hat or outfit on the person in your video.** Video on, as swap the person. Picture 1 is a still of them from the video (a pasted screenshot works), as person. Picture 2 is the hat or outfit, as outfit: `@1 wears @2 in place of his cap`\\n- **A longer story.** Video on, as keep going. Load your last clip and type what comes next.\", \"models\": [], \"packs\": [], \"links\": []}"
    },
    "color": "#14424d",
    "bgcolor": "#161f21"
@@ -3581,10 +3581,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "{\"v\": 1, \"title\": \"Your video and sound\", \"subtitle\": \"Up to 15 seconds. Pick its job on H3 Prompt Writer.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"Drag the handles to the part you want. The clip takes this video's shape, and takes about three times as long to make.\\n\\n- **copy its moves**: your person does what the video shows. The clothes in the video stay unless your line names new ones.\\n- **swap the person**: the same video with your person in it. Type nothing. Two people in it? Type who goes: `the child in the seat`.\\n- **keep going**: more of the clip, from where it ends. Type what comes next. It saves the new part.\\n- **new voice**: the same video in another voice. Turn **Voice** on and put the words in quotes.\\n\\n### Voice and Song\\n**Voice**: about 10 seconds of someone talking, yours or one you have permission to use. Put the words in quotes, on the last line.\\n\\n**Song**: it plays as it is, and each moment moves onto its beat. With a voice too, it goes under the words.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "{\"v\": 1, \"title\": \"Your video and sound\", \"subtitle\": \"Up to 15 seconds. Pick its job on H3 Prompt Writer.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"Drag the handles to the part you want. The clip takes this video's shape, and takes about three times as long to make.\\n\\n- **copy its moves**: your person does what the video shows. The clothes in the video stay unless your line names new ones.\\n- **swap the person**: the same video with your person in it. Type nothing. Two people in it? Type who goes: `the child in the seat`. Same person in a new hat or outfit? Load a still of them as person and the item as outfit (Recipes).\\n- **keep going**: more of the clip, from where it ends. Type what comes next. It saves the new part.\\n- **new voice**: the same video in another voice. Turn **Voice** on and put the words in quotes.\\n\\n### Voice and Song\\n**Voice**: about 10 seconds of someone talking, yours or one you have permission to use. Put the words in quotes, on the last line.\\n\\n**Song**: it plays as it is, and each moment moves onto its beat. With a voice too, it goes under the words.\", \"models\": [], \"packs\": [], \"links\": []}"
    ],
    "widgets_values_named": {
-    "note": "{\"v\": 1, \"title\": \"Your video and sound\", \"subtitle\": \"Up to 15 seconds. Pick its job on H3 Prompt Writer.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"Drag the handles to the part you want. The clip takes this video's shape, and takes about three times as long to make.\\n\\n- **copy its moves**: your person does what the video shows. The clothes in the video stay unless your line names new ones.\\n- **swap the person**: the same video with your person in it. Type nothing. Two people in it? Type who goes: `the child in the seat`.\\n- **keep going**: more of the clip, from where it ends. Type what comes next. It saves the new part.\\n- **new voice**: the same video in another voice. Turn **Voice** on and put the words in quotes.\\n\\n### Voice and Song\\n**Voice**: about 10 seconds of someone talking, yours or one you have permission to use. Put the words in quotes, on the last line.\\n\\n**Song**: it plays as it is, and each moment moves onto its beat. With a voice too, it goes under the words.\", \"models\": [], \"packs\": [], \"links\": []}"
+    "note": "{\"v\": 1, \"title\": \"Your video and sound\", \"subtitle\": \"Up to 15 seconds. Pick its job on H3 Prompt Writer.\", \"author\": \"AusBoss\", \"layout\": \"card\", \"accent\": \"\", \"body\": \"Drag the handles to the part you want. The clip takes this video's shape, and takes about three times as long to make.\\n\\n- **copy its moves**: your person does what the video shows. The clothes in the video stay unless your line names new ones.\\n- **swap the person**: the same video with your person in it. Type nothing. Two people in it? Type who goes: `the child in the seat`. Same person in a new hat or outfit? Load a still of them as person and the item as outfit (Recipes).\\n- **keep going**: more of the clip, from where it ends. Type what comes next. It saves the new part.\\n- **new voice**: the same video in another voice. Turn **Voice** on and put the words in quotes.\\n\\n### Voice and Song\\n**Voice**: about 10 seconds of someone talking, yours or one you have permission to use. Put the words in quotes, on the last line.\\n\\n**Song**: it plays as it is, and each moment moves onto its beat. With a voice too, it goes under the words.\", \"models\": [], \"packs\": [], \"links\": []}"
    },
    "color": "#14424d",
    "bgcolor": "#161f21"
@@ -3738,7 +3738,7 @@
      "switches": [
       {
        "id": 5,
-       "title": "Video: swap or copy moves",
+       "title": "Video: your clip",
        "nodes": [
         150
        ],
@@ -3850,7 +3850,7 @@
          "fast test": false
         },
         "150": {
-         "megapixels": 1
+         "megapixels": 0.41
         }
        }
       }
@@ -3987,10 +3987,10 @@
     "ver": "2.21.0"
    },
    "widgets_values": [
-    "\u2b05\ufe0f#224 **Video**: swap someone in or copy the moves. Up to 10 s, at about 480p.\n\u2b05\ufe0f#224 **Keep the clip's sound**: its own words and music, for a swap.\n\u2198\ufe0f#151 **Voice**: 10 seconds of talking, with the words in quotes on the last line.\n\u2198\ufe0f#152 **Song**: plays under the clip."
+    "\u2b05\ufe0f#224 **Video**: swap someone in, change what they wear, copy the moves or keep it going. Up to 10 s, at about 480p.\n\u2b05\ufe0f#224 **Keep the clip's sound**: its own words and music, for a swap or a new outfit.\n\u2198\ufe0f#151 **Voice**: 10 seconds of talking, with the words in quotes on the last line.\n\u2198\ufe0f#152 **Song**: plays under the clip."
    ],
    "widgets_values_named": {
-    "text": "\u2b05\ufe0f#224 **Video**: swap someone in or copy the moves. Up to 10 s, at about 480p.\n\u2b05\ufe0f#224 **Keep the clip's sound**: its own words and music, for a swap.\n\u2198\ufe0f#151 **Voice**: 10 seconds of talking, with the words in quotes on the last line.\n\u2198\ufe0f#152 **Song**: plays under the clip."
+    "text": "\u2b05\ufe0f#224 **Video**: swap someone in, change what they wear, copy the moves or keep it going. Up to 10 s, at about 480p.\n\u2b05\ufe0f#224 **Keep the clip's sound**: its own words and music, for a swap or a new outfit.\n\u2198\ufe0f#151 **Voice**: 10 seconds of talking, with the words in quotes on the last line.\n\u2198\ufe0f#152 **Song**: plays under the clip."
    },
    "color": "#07090a",
    "bgcolor": "#07090a"
@@ -5207,7 +5207,7 @@
       ]
      },
      {
-      "id": "f14285c7-e429-4811-9dce-ac349af3a4e5",
+      "id": "f836716c-c34c-4248-8c23-d60864273611",
       "name": "clip sound",
       "type": "AUDIO",
       "linkIds": [
@@ -5219,7 +5219,7 @@
       ]
      },
      {
-      "id": "d619bfaa-dc51-4cbd-a4b1-e5822815a73c",
+      "id": "dfc070d2-400b-4f79-98da-7844a3cd1818",
       "name": "steer mouths",
       "type": "BOOLEAN",
       "linkIds": [
@@ -7937,10 +7937,10 @@
        "ver": "2.21.0"
       },
       "widgets_values": [
-       "floor(a * c * (1 - 0.375 * b) / 16) * 16"
+       "floor(a * c * (1 - 0.375 * b) / 32) * 32"
       ],
       "widgets_values_named": {
-       "expression": "floor(a * c * (1 - 0.375 * b) / 16) * 16"
+       "expression": "floor(a * c * (1 - 0.375 * b) / 32) * 32"
       },
       "color": "#14424d",
       "bgcolor": "#161f21"
@@ -8014,10 +8014,10 @@
        "ver": "2.21.0"
       },
       "widgets_values": [
-       "floor(a * c * (1 - 0.375 * b) / 16) * 16"
+       "floor(a * c * (1 - 0.375 * b) / 32) * 32"
       ],
       "widgets_values_named": {
-       "expression": "floor(a * c * (1 - 0.375 * b) / 16) * 16"
+       "expression": "floor(a * c * (1 - 0.375 * b) / 32) * 32"
       },
       "color": "#14424d",
       "bgcolor": "#161f21"

nodes/_h3_cast_helpers.py

@@ -37,6 +37,12 @@ NEEDS_A_PERSON = {
     MOVES: "To copy the moves, connect a picture of your person and pick person for it.",
     SWAP: "To swap the person, connect a picture of the new person and pick person for it.",
 }
+# A hat or a shirt with no person picture: the user wants the video's own person to wear it.
+DRESS_THE_VIDEO_PERSON = (
+    "To put this on the person in your video, load a still of them from the video as Picture 1 "
+    "and pick person for it. Give the hat or outfit its own picture, pick outfit or thing, and type "
+    "\"@1 wears @2\" with video job swap the person."
+)
 
 # (what the label is, how it is kept). {picture} is H3's number for the picture.
 JOB_WORDS = {
@@ -241,6 +247,8 @@ class Cast:
         if not self.pictures and not self.video:
             return NOTHING_IN
         if self.video in NEEDS_A_PERSON and not self.people:
+            if self.of(OUTFIT, THING):
+                return DRESS_THE_VIDEO_PERSON
             return NEEDS_A_PERSON[self.video]
         typed = sorted({int(number) for number in TOKEN.findall(text)} - {0})
         slots = {picture.slot: picture for picture in self.pictures}

scripts/workflow_contract.py

@@ -63,6 +63,58 @@ def named_copy_problems(node: dict) -> list[str]:
     return [f"node {node['id']} has stale named widget values: {', '.join(stale)}"] if stale else []
 
 
+def saved_widget(node: dict, name: str):
+    """A widget's saved value by name, or KeyError when the node keeps no named copy of it."""
+    values = node.get("widgets_values")
+    if isinstance(values, dict) and name in values:
+        return values[name]
+    named = node.get("widgets_values_named")
+    if isinstance(named, dict) and name in named:
+        return named[name]
+    raise KeyError(name)
+
+
+def switch_row_problems(card: dict, by_id: dict) -> list[str]:
+    """A Workflow Switches row opens striped "changed", with neither off nor
+    on picked, when the values it sets are saved half one way and half the
+    other, or when its nodes run one way and its values say the other. Every
+    row of an example has to open plainly off or on."""
+    settings = (card.get("properties") or {}).get("ausboss_workflow_switches") or {}
+    problems = []
+    for row in settings.get("switches") or []:
+        title = row.get("title", row.get("id"))
+        sides = set()
+        on, off = row.get("on") or {}, row.get("off") or {}
+        for node_id, widgets in on.items():
+            node = by_id.get(int(node_id)) if str(node_id).isdigit() else None
+            if node is None or not isinstance(widgets, dict):
+                continue
+            for name, on_value in widgets.items():
+                off_value = (off.get(node_id) or {}).get(name)
+                lora_rows = isinstance(on_value, str) and on_value.startswith("[")
+                if isinstance(on_value, (dict, list)) or lora_rows or on_value == off_value:
+                    continue  # a LoRA stack is matched row by row, and a shared value says nothing
+                try:
+                    saved = saved_widget(node, name)
+                except KeyError:
+                    continue
+                if saved == on_value:
+                    sides.add("on")
+                elif saved == off_value:
+                    sides.add("off")
+                else:
+                    sides.add(f"neither ({node_id} {name} is {saved!r}, not {on_value!r} or {off_value!r})")
+        members = [by_id[i] for i in row.get("nodes") or [] if i in by_id]
+        modes = {"on" if m.get("mode", 0) == 0 else "off" for m in members}
+        if len(modes) > 1:
+            problems.append(f"switch card {card['id']} row '{title}': some of its nodes are on and some off")
+        odd = sorted(side for side in sides if side not in ("on", "off"))
+        if odd or len(sides) > 1 or (sides and modes and sides != modes):
+            detail = "; ".join(odd) if odd else f"values say {'/'.join(sorted(sides))}, nodes say {'/'.join(sorted(modes)) or '-'}"
+            problems.append(f"switch card {card['id']} row '{title}' opens as changed: {detail}")
+    return problems
+
+
 def workflow_problems(data: dict) -> list[str]:
     problems = []
     nodes = data.get("nodes", [])
@@ -112,6 +164,8 @@ def workflow_problems(data: dict) -> list[str]:
         problems.append(f"saved zoom {scale} is below {MIN_ZOOM}; widget text would not draw")
     for n in nodes:
         problems.extend(named_copy_problems(n))
+        if n["type"] == "AUSBOSS_NODES_WorkflowSwitches":
+            problems.extend(switch_row_problems(n, by_id))
         if n["type"] in CORE_LORA_LOADERS:
             problems.append(f"node {n['id']} loads a LoRA with core {n['type']}; examples use LoRA Loader 🆎")
         selected = selected_files(n)

tests/test_h3_prompt_writer.py

@@ -652,6 +652,12 @@ class VideoJobTests(unittest.TestCase):
         cast = Cast(number_pictures({1: "place"}, "copy its moves"), "copy its moves")
         self.assertIn("To copy the moves", cast.problem("She dances."))
 
+    def test_a_hat_with_no_person_picture_says_how_to_put_it_on_the_video_person(self):
+        for job in ("outfit", "thing"):
+            for video in ("swap the person", "copy its moves"):
+                cast = Cast(number_pictures({1: job}, video), video)
+                self.assertIn("load a still of them from the video as Picture 1", cast.problem("He wears @1."))
+
     def test_a_swap_with_nobody_to_put_in_says_so(self):
         with self.assertRaisesRegex(ValueError, "new person"):
             write_prompt("", jobs={}, video="swap the person")

tests/test_workflow_contract.py

@@ -38,6 +38,24 @@ class WorkflowContractTests(unittest.TestCase):
         graph["nodes"][0]["widgets_values_named"] = {"video": "someone_else.mp4", "start_seconds": 0, "upload": ""}
         self.assertIn("stale named widget values: video", " ".join(workflow_problems(graph)))
 
+    def test_a_switch_row_must_open_plainly_off_or_on(self):
+        graph = good_graph()
+        graph["nodes"][0]["widgets_values"] = {"megapixels": 0.41}
+        card = {"id": 4, "type": "AUSBOSS_NODES_WorkflowSwitches", "pos": [0, 300], "size": [100, 40], "properties": {
+            "ausboss_workflow_switches": {"switches": [{"id": 9, "title": "Fast test", "nodes": [],
+                "on": {"1": {"megapixels": 0.26}}, "off": {"1": {"megapixels": 1}}}]}}}
+        graph["nodes"].append(card)
+        graph["last_node_id"] = 4
+        graph["groups"][0]["bounding"] = [-20, 40, 320, 320]
+        self.assertIn("row 'Fast test' opens as changed", " ".join(workflow_problems(graph)))
+        card["properties"]["ausboss_workflow_switches"]["switches"][0]["off"] = {"1": {"megapixels": 0.41}}
+        self.assertEqual(workflow_problems(graph), [])
+        graph["nodes"][1]["widgets_values"] = {"steps": 8}
+        graph["nodes"][1]["mode"] = 4
+        card["properties"]["ausboss_workflow_switches"]["switches"].append(
+            {"id": 10, "title": "Turbo", "nodes": [2], "on": {"2": {"steps": 8}}, "off": {"2": {"steps": 20}}})
+        self.assertIn("row 'Turbo' opens as changed: values say on, nodes say off", " ".join(workflow_problems(graph)))
+
     def test_workflow_id_and_zoom(self):
         graph = good_graph()
         graph["id"] = "3f2b8c1e-5d4a-4e6f-9a7b-1c2d3e4f5a6b"