workflows comfy

docs/serviceTemplates/ltx-video-ugc-multishot.json

@@ -0,0 +1,107 @@
+{
+  "id": "ltx-video-ugc-multishot",
+  "name": "ComfyUI — UGC multishot reel (LTX-2.3)",
+  "description": "ComfyUI preloaded with LTX-2.3 for a vertical UGC reel built one shot at a time, with the same character in every shot. It runs the same two-stage generator as the product-video template (22B dev fp8, base pass then a spatial-upscaler refine pass), so shot quality matches. Usage: optionally upload a character photo as the first frame — with no photo you turn on 'No character image' and the style prompt builds the character instead. Then fill two text boxes: 'Style — every shot' holds everything that must stay identical (who the character is, wardrobe, room, lighting, palette, camera look) and you never touch it again; 'Scene — this shot' holds only what changes (action, camera move, dialogue, ambient sound). Click Run for a clip, edit the scene box, Run again. Every run also saves that clip's last frame beside it as lastframe_….png, and a Shot start dial decides how it is used: at 1.0 it is frame 0 and the shot continues seamlessly, while around 0.35 it acts as a character reference only, so the model composes a new frame and you get a hard cut to a new scene with the same character. Chaining shots this way — shot 2 then starts exactly where shot 1 ended and identity carries in pixels rather than in prose, the trick that holds a multi-shot reel together. Clips are 9:16 vertical (720×1280) at 5 seconds each (126 frames at 25 fps) with audio, saved to the bucket root; ComfyUI's save counter never overwrites, so re-takes accumulate and you pick the good ones later. Once every shot is good, switch to the Assemble reel workflow to concatenate up to 8 of them into one clip inside ComfyUI — no external editor needed. The assemble workflow keeps each clip's own audio and runs it through a voice-conversion node: LTX re-rolls the speaker on every clip, so the stitched track would otherwise change voice at each cut. It takes the reference voice from Shot 1's own audio — nothing to upload and nothing to type, since this is speech-to-speech rather than transcription — and converts every line to that one speaker, with words, timing and room tone untouched — so the picture stays in sync and the lips still match. The node pack is not in the image; the bootstrap clones it and installs its dependencies on launch, but only when the graph actually uses it, so the first launch of this template is slower and the product template is unaffected. Select a persistent-storage bucket: it holds ComfyUI's whole base directory, so the 38 GiB of weights download once and are reused, and the numbered shot clips land in the bucket root where the storage API's listFiles can see them. Needs a CUDA GPU with 48 GB+ VRAM.",
+  "kind": "bundle",
+  "service": "comfyui",
+  "outcome": "Build a multi-shot 9:16 UGC reel with one character and one voice throughout.",
+  "category": "video",
+  "includes": [
+    {
+      "name": "LTX-2.3 22B dev (fp8)",
+      "kind": "model",
+      "sizeGb": 29.2,
+      "repoId": "Lightricks/LTX-2.3-fp8"
+    },
+    {
+      "name": "Gemma-3-12B-it text encoder (fp4 mixed)",
+      "kind": "model",
+      "sizeGb": 9.5,
+      "repoId": "Comfy-Org/ltx-2"
+    },
+    {
+      "name": "LTX-2.3 22B distilled LoRA (rank 111)",
+      "kind": "model",
+      "sizeGb": 2.7,
+      "repoId": "Comfy-Org/ltx-2.3"
+    },
+    {
+      "name": "LTX-2.3 spatial upscaler x2",
+      "kind": "model",
+      "sizeGb": 1.0,
+      "repoId": "Lightricks/LTX-2.3"
+    },
+    {
+      "name": "Gemma-3-12B-it abliterated LoRA (rank 64)",
+      "kind": "model",
+      "sizeGb": 0.6,
+      "repoId": "Comfy-Org/ltx-2"
+    },
+    {
+      "name": "TTS-Audio-Suite (voice conversion for the assemble workflow)",
+      "kind": "customnode",
+      "url": "https://github.com/diodiogod/TTS-Audio-Suite"
+    }
+  ],
+  "image": "yanwk/comfyui-boot",
+  "tag": "cu130-megapak-pt211-20260803",
+  "exposedPorts": [
+    8188
+  ],
+  "entrypoint": [
+    "/bin/bash",
+    "-c"
+  ],
+  "commandFile": "comfyui-ugc-bootstrap.sh",
+  "workflows": [
+    {
+      "id": "ocean_ugc_multishot",
+      "name": "Multishot reel — one shot per Run",
+      "description": "An image input plus two prompt boxes — a style box reused by every shot and a scene box you retype between runs. One Run renders one 9:16 clip and saves its last frame as lastframe_….png. The Shot start dial decides whether that image continues the take seamlessly (1.0) or acts as a character reference for a hard cut to a new scene (~0.35). With no image at all, the style prompt builds the character instead.",
+      "file": "workflows/ocean_ugc_multishot.json"
+    },
+    {
+      "id": "ocean_ugc_assemble",
+      "name": "Assemble reel",
+      "description": "Concatenates up to 8 rendered shot clips (video + audio) into a single reel using core ComfyUI video nodes, then converts every voice in the stitched track to Shot 1's voice so the reel does not change speaker at each cut — pick clips in the LoadVideo dropdowns, bypass any slots you don't need, and Run. No upload and no script: it reads the reference voice straight off the clips you already rendered.",
+      "file": "workflows/ocean_ugc_assemble.json"
+    }
+  ],
+  "userConfigurableEnvVars": [
+    {
+      "key": "COMFY_WORKFLOW_ID",
+      "validation": "^[A-Za-z0-9_-]+$"
+    },
+    {
+      "key": "COMFY_WORKFLOW"
+    }
+  ],
+  "requiredResources": [
+    {
+      "id": "cpu",
+      "min": 8,
+      "recommended": 16,
+      "unit": "cores"
+    },
+    {
+      "id": "ram",
+      "min": 48,
+      "recommended": 128,
+      "unit": "GB"
+    },
+    {
+      "id": "disk",
+      "min": 45,
+      "recommended": 85,
+      "unit": "GB"
+    },
+    {
+      "kind": "discrete",
+      "type": "gpu",
+      "min": 1,
+      "recommended": 1,
+      "unit": "count",
+      "description": "CUDA GPU, 48 GB+ VRAM recommended (LTX-2.3 22B fp8 + Gemma-3-12B encoder)"
+    }
+  ]
+}

docs/serviceTemplates/ltx-video-ugc-product.json

@@ -0,0 +1,96 @@
+{
+  "id": "ltx-video-ugc-product",
+  "name": "ComfyUI — UGC product video (LTX-2.3)",
+  "description": "ComfyUI preloaded with LTX-2.3 for vertical short-form product video: a product photo becomes a 9:16 clip with camera motion and ambient audio, capped at the graph's 5 seconds (126 frames at 25 fps). Select a persistent-storage bucket: it holds ComfyUI's whole base directory, so the 38 GiB of weights download once and are reused, and generated clips land in the bucket root. Needs a CUDA GPU with 48 GB+ VRAM.",
+  "kind": "bundle",
+  "service": "comfyui",
+  "outcome": "Turn a product photo into a 9:16 clip with camera motion and ambient audio.",
+  "category": "video",
+  "includes": [
+    {
+      "name": "LTX-2.3 22B dev (fp8)",
+      "kind": "model",
+      "sizeGb": 29.2,
+      "repoId": "Lightricks/LTX-2.3-fp8"
+    },
+    {
+      "name": "Gemma-3-12B-it text encoder (fp4 mixed)",
+      "kind": "model",
+      "sizeGb": 9.5,
+      "repoId": "Comfy-Org/ltx-2"
+    },
+    {
+      "name": "LTX-2.3 22B distilled LoRA (rank 111)",
+      "kind": "model",
+      "sizeGb": 2.7,
+      "repoId": "Comfy-Org/ltx-2.3"
+    },
+    {
+      "name": "LTX-2.3 spatial upscaler x2",
+      "kind": "model",
+      "sizeGb": 1.0,
+      "repoId": "Lightricks/LTX-2.3"
+    },
+    {
+      "name": "Gemma-3-12B-it abliterated LoRA (rank 64)",
+      "kind": "model",
+      "sizeGb": 0.6,
+      "repoId": "Comfy-Org/ltx-2"
+    }
+  ],
+  "image": "yanwk/comfyui-boot",
+  "tag": "cu130-megapak-pt211-20260803",
+  "exposedPorts": [
+    8188
+  ],
+  "entrypoint": [
+    "/bin/bash",
+    "-c"
+  ],
+  "commandFile": "comfyui-ugc-bootstrap.sh",
+  "workflows": [
+    {
+      "id": "ocean_ugc_product",
+      "name": "Product → vertical clip",
+      "description": "Upload a product photo, get a 9:16 clip with camera motion and ambient audio.",
+      "file": "workflows/ocean_ugc_product.json"
+    }
+  ],
+  "userConfigurableEnvVars": [
+    {
+      "key": "COMFY_WORKFLOW_ID",
+      "validation": "^[A-Za-z0-9_-]+$"
+    },
+    {
+      "key": "COMFY_WORKFLOW"
+    }
+  ],
+  "requiredResources": [
+    {
+      "id": "cpu",
+      "min": 8,
+      "recommended": 16,
+      "unit": "cores"
+    },
+    {
+      "id": "ram",
+      "min": 48,
+      "recommended": 128,
+      "unit": "GB"
+    },
+    {
+      "id": "disk",
+      "min": 40,
+      "recommended": 80,
+      "unit": "GB"
+    },
+    {
+      "kind": "discrete",
+      "type": "gpu",
+      "min": 1,
+      "recommended": 1,
+      "unit": "count",
+      "description": "CUDA GPU, 48 GB+ VRAM recommended (LTX-2.3 22B fp8 + Gemma-3-12B encoder)"
+    }
+  ]
+}