{ "id": "e8f30439-267f-45de-b260-db0d114b2c85", "revision": 0, "last_node_id": 24, "last_link_id": 52, "nodes": [ { "id": 9, "type": "SaveImage", "pos": [ 840, 0 ], "size": [ 570, 700 ], "flags": {}, "order": 4, "mode": 0, "inputs": [ { "name": "images", "type": "IMAGE", "link": 22 } ], "outputs": [ { "name": "images", "type": "IMAGE", "links": null } ], "properties": { "Node name for S&R": "SaveImage" }, "widgets_values": [ "moss_image_t2i" ], "widgets_values_named": { "filename_prefix": "moss_image_t2i" } }, { "id": 18, "type": "3301b744-2960-415a-83b0-fe8a89d99fdb", "pos": [ 360, 0 ], "size": [ 430, 680 ], "flags": {}, "order": 3, "mode": 0, "inputs": [ { "label": "positive_prompt", "name": "value", "type": "STRING", "widget": { "name": "value" }, "link": null }, { "label": "enable_prompt_upsampler", "name": "switch", "type": "BOOLEAN", "widget": { "name": "switch" }, "link": null }, { "label": "negative_prompt", "name": "text", "type": "STRING", "widget": { "name": "text" }, "link": null }, { "name": "width", "type": "INT", "widget": { "name": "width" }, "link": 40 }, { "name": "height", "type": "INT", "widget": { "name": "height" }, "link": 41 }, { "label": "SLG", "name": "switch_1", "type": "BOOLEAN", "widget": { "name": "switch_1" }, "link": null }, { "label": "unet_model", "name": "unet_name", "type": "COMBO", "widget": { "name": "unet_name" }, "link": null }, { "label": "clip_model", "name": "clip_name", "type": "COMBO", "widget": { "name": "clip_name" }, "link": null }, { "label": "vae_model", "name": "vae_name", "type": "COMBO", "widget": { "name": "vae_name" }, "link": null } ], "outputs": [ { "name": "IMAGE", "type": "IMAGE", "links": [ 22 ] } ], "properties": { "previewExposures": [] }, "widgets_values": [ "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle", true, "nsfw, worst quality, bad anatomy, watermark, jpeg artifact", 1024, 1024, 333599710072681, 25, 2.5, "er_sde", "simple", false, "moss_image_sft_int8_convrot.safetensors", "qwen3.5_9b_w4a8_gptq.safetensors", "flux2-vae_small.safetensors" ], "widgets_values_named": { "value": "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle", "switch": true, "text": "nsfw, worst quality, bad anatomy, watermark, jpeg artifact", "width": 1024, "height": 1024, "seed": 333599710072681, "steps": 25, "cfg": 2.5, "sampler_name": "er_sde", "scheduler": "simple", "switch_1": false, "unet_name": "moss_image_sft_int8_convrot.safetensors", "clip_name": "qwen3.5_9b_w4a8_gptq.safetensors", "vae_name": "flux2-vae_small.safetensors" } }, { "id": 19, "type": "ResolutionSelector", "pos": [ 10, 350 ], "size": [ 280, 190 ], "flags": {}, "order": 0, "mode": 0, "showAdvanced": true, "inputs": [], "outputs": [ { "name": "width", "type": "INT", "links": [ 40 ] }, { "name": "height", "type": "INT", "links": [ 41 ] } ], "properties": { "Node name for S&R": "ResolutionSelector" }, "widgets_values": [ "1:1 (Square)", 1, 16 ], "widgets_values_named": { "aspect_ratio": "1:1 (Square)", "megapixels": 1, "multiple": 16 } }, { "id": 21, "type": "MarkdownNote", "pos": [ -1210, 10 ], "size": [ 560, 1000 ], "flags": {}, "order": 1, "mode": 0, "inputs": [], "outputs": [], "properties": {}, "widgets_values": [ "Guide: [Subgraph](https://docs.comfy.org/interface/features/subgraph)\n\n## Model Links\n\n- [Hugging Face:Comfy-Org/MOSS-Image](https://huggingface.co/Comfy-Org/MOSS-Image)\n\n**diffusion_models**\n\n- [moss_image_sft_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_int8_convrot.safetensors)\n- [moss_image_sft_bf16.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_bf16.safetensors)\n\n**text_encoders**\n\n- [qwen3.5_9b_w4a8_gptq.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/text_encoders/qwen3.5_9b_w4a8_gptq.safetensors)\n\n**Checkpoint variants**: `moss_image_sft_*` is the instruction-tuned checkpoint used here; `moss_image_base_*` runs in the same graph.\n\n**vae**\n\n- [flux2-vae_small.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/vae/flux2-vae_small.safetensors)\n\n\n## Model Storage Location\n\n```\nšŸ“‚ ComfyUI/\nā”œā”€ā”€ šŸ“‚ models/\nā”‚ ā”œā”€ā”€ šŸ“‚ diffusion_models/\nā”‚ ā”‚ ā”œā”€ā”€ moss_image_sft_int8_convrot.safetensors\nā”‚ ā”‚ ā””ā”€ā”€ moss_image_sft_bf16.safetensors\nā”‚ ā”œā”€ā”€ šŸ“‚ text_encoders/\nā”‚ ā”‚ ā””ā”€ā”€ qwen3.5_9b_w4a8_gptq.safetensors\nā”‚ ā””ā”€ā”€ šŸ“‚ vae/\nā”‚ ā””ā”€ā”€ flux2-vae_small.safetensors\n```\n\n## Prompt upsampler\n\nThe `enable_prompt_upsampler` switch on the **Text to Image (MOSS-Image)** node is **true**, so a short prompt is rewritten into one full\nscene caption before it reaches the model. The upsampler runs on MOSS-Image's own text encoder\n(`qwen3.5_9b_w4a8_gptq.safetensors`, loaded with `type = moss_image`), the same file the model already needs ā€” no second\ntext encoder to download. Set the switch to `false` to send the raw prompt; either way the caption actually used is shown in the\n`PreviewAny` node inside the subgraph.\n\n\n## Prompt skill\n\nThe system prompt driving the upsampler is written from the official [MOSS-Image Prompt Enhancer skill](https://github.com/OpenMOSS/MOSS-Image/blob/main/skills/moss-image-prompt-enhancer/SKILL.md). The same skill turns an idea into a caption or a Layout prompt outside ComfyUI, and it is the reference for the caption style this model was trained on.\n\n## Report Issue\n\nNote: Please update ComfyUI first ([guide](https://docs.comfy.org/installation/update_comfyui)) and prepare required models. Desktop/Cloud updates follow stable releases, so some nightly-supported models may not be available yet.\n\n- Cannot run / runtime errors: [ComfyUI/issues](https://github.com/comfyanonymous/ComfyUI/issues)\n- UI / frontend issues: [ComfyUI_frontend/issues](https://github.com/Comfy-Org/ComfyUI_frontend/issues)\n- Workflow issues: [workflow_templates/issues](https://github.com/Comfy-Org/workflow_templates/issues)\n" ], "widgets_values_named": { "text": "Guide: [Subgraph](https://docs.comfy.org/interface/features/subgraph)\n\n## Model Links\n\n- [Hugging Face:Comfy-Org/MOSS-Image](https://huggingface.co/Comfy-Org/MOSS-Image)\n\n**diffusion_models**\n\n- [moss_image_sft_int8_convrot.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_int8_convrot.safetensors)\n- [moss_image_sft_bf16.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_bf16.safetensors)\n\n**text_encoders**\n\n- [qwen3.5_9b_w4a8_gptq.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/text_encoders/qwen3.5_9b_w4a8_gptq.safetensors)\n\n**Checkpoint variants**: `moss_image_sft_*` is the instruction-tuned checkpoint used here; `moss_image_base_*` runs in the same graph.\n\n**vae**\n\n- [flux2-vae_small.safetensors](https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/vae/flux2-vae_small.safetensors)\n\n\n## Model Storage Location\n\n```\nšŸ“‚ ComfyUI/\nā”œā”€ā”€ šŸ“‚ models/\nā”‚ ā”œā”€ā”€ šŸ“‚ diffusion_models/\nā”‚ ā”‚ ā”œā”€ā”€ moss_image_sft_int8_convrot.safetensors\nā”‚ ā”‚ ā””ā”€ā”€ moss_image_sft_bf16.safetensors\nā”‚ ā”œā”€ā”€ šŸ“‚ text_encoders/\nā”‚ ā”‚ ā””ā”€ā”€ qwen3.5_9b_w4a8_gptq.safetensors\nā”‚ ā””ā”€ā”€ šŸ“‚ vae/\nā”‚ ā””ā”€ā”€ flux2-vae_small.safetensors\n```\n\n## Prompt upsampler\n\nThe `enable_prompt_upsampler` switch on the **Text to Image (MOSS-Image)** node is **true**, so a short prompt is rewritten into one full\nscene caption before it reaches the model. The upsampler runs on MOSS-Image's own text encoder\n(`qwen3.5_9b_w4a8_gptq.safetensors`, loaded with `type = moss_image`), the same file the model already needs ā€” no second\ntext encoder to download. Set the switch to `false` to send the raw prompt; either way the caption actually used is shown in the\n`PreviewAny` node inside the subgraph.\n\n\n## Prompt skill\n\nThe system prompt driving the upsampler is written from the official [MOSS-Image Prompt Enhancer skill](https://github.com/OpenMOSS/MOSS-Image/blob/main/skills/moss-image-prompt-enhancer/SKILL.md). The same skill turns an idea into a caption or a Layout prompt outside ComfyUI, and it is the reference for the caption style this model was trained on.\n\n## Report Issue\n\nNote: Please update ComfyUI first ([guide](https://docs.comfy.org/installation/update_comfyui)) and prepare required models. Desktop/Cloud updates follow stable releases, so some nightly-supported models may not be available yet.\n\n- Cannot run / runtime errors: [ComfyUI/issues](https://github.com/comfyanonymous/ComfyUI/issues)\n- UI / frontend issues: [ComfyUI_frontend/issues](https://github.com/Comfy-Org/ComfyUI_frontend/issues)\n- Workflow issues: [workflow_templates/issues](https://github.com/Comfy-Org/workflow_templates/issues)\n" }, "color": "#222", "bgcolor": "#000" }, { "id": 23, "type": "MarkdownNote", "pos": [ -600, 10 ], "size": [ 520, 1200 ], "flags": {}, "order": 2, "mode": 0, "inputs": [], "outputs": [], "properties": {}, "widgets_values": [ "## MOSS-Image capabilities and limits\n\nSources: [USERGUIDE](https://github.com/OpenMOSS/MOSS-Image/blob/main/USERGUIDE.md), README, HF Base/SFT cards; cross-checked against community benchmark notes.\n\n### Resolution\n\n| Item | Official |\n|------|----------|\n| Recommended total pixels | **1M, 1.5M, or 2M** |\n| Extrapolation range | **0.25M (512Ɨ512) to 4M (~2K)** |\n| Aspect ratio | Arbitrary; width/height must be positive multiples of **16** |\n| This template default | 1024Ɨ1024 (~1M) |\n\n### Strengths\n\n- ~**5B** single-stream DiT (Muon); FLUX.2 small VAE; Qwen-family text encoder.\n- Strong bilingual English and Chinese text rendering and auto layout; JSON Layout control for text/objects; long prompts (~**4K** tokens); anime and photorealistic.\n- **SFT** clearer text grouping/layouts; **Base** has no post-training / CFG distillation (better for community LoRAs).\n- Inputs: natural language, tag sequences, JSON layout (layout is experimental). Prefer English (most training data); Chinese supported.\n\n### Limits / notes\n\n- Auto line breaks may not match intent ā€” use `\\n` or real newlines for precise text.\n- Turning off SLG looks more realistic but can collapse details; official negative-branch skip layer **22** (24/26 also help). kijai notes SLG can overblow anime styles.\n- Recommended negative prompt: `nsfw, worst quality, bad anatomy, watermark, jpeg artifact`\n- If VRAM is tight, lower resolution or use Sequential / CPU offload.\n- When training LoRA yourself, **freeze the text encoder**.\n- License: Apache 2.0.\n\n---\n\n## Sampling settings vs. official defaults\n\n| Setting | This template | OpenMOSS official |\n|---------|---------------|-------------------|\n| Steps | 25 | 50 |\n| CFG | 2.5 | 3 |\n| Sampler | er_sde | Euler |\n| Scheduler | simple | default |\n| Denoise | 1 | 1 |\n| Shift | 5 (model default) | 5 (same) |\n\nThis template follows the faster settings from kijai's ComfyUI MOSS-Image PR [#17095](https://github.com/comfyanonymous/ComfyUI/pull/17095). That is a speed trade-off, not a bug. For results closer to the official ones, try 50 steps and CFG 3.\n\n### Community notes\n\n- **Skip layer guidance:** can overblow anime styles; no alternative values were suggested (template keeps official layer 22). You can check the detail settings inside the subgraph\n- **Shift:** adjust with ModelSamplingAuraFlow or ModelSamplingFlux ā€” ModelSamplingSD3 uses a completely different scale. You can enter the subgraph to adjust it\n- **SFT banding:** experimental LoRA at about āˆ’0.5 strength: [moss_image_sft_lora_blocks_21_to_49ā€¦](https://huggingface.co/Kijai/MOSS-image_experimental/blob/main/loras/moss_image_sft_lora_blocks_21_to_49_rank_512_bf16.safetensors)\n" ], "widgets_values_named": { "text": "## MOSS-Image capabilities and limits\n\nSources: [USERGUIDE](https://github.com/OpenMOSS/MOSS-Image/blob/main/USERGUIDE.md), README, HF Base/SFT cards; cross-checked against community benchmark notes.\n\n### Resolution\n\n| Item | Official |\n|------|----------|\n| Recommended total pixels | **1M, 1.5M, or 2M** |\n| Extrapolation range | **0.25M (512Ɨ512) to 4M (~2K)** |\n| Aspect ratio | Arbitrary; width/height must be positive multiples of **16** |\n| This template default | 1024Ɨ1024 (~1M) |\n\n### Strengths\n\n- ~**5B** single-stream DiT (Muon); FLUX.2 small VAE; Qwen-family text encoder.\n- Strong bilingual English and Chinese text rendering and auto layout; JSON Layout control for text/objects; long prompts (~**4K** tokens); anime and photorealistic.\n- **SFT** clearer text grouping/layouts; **Base** has no post-training / CFG distillation (better for community LoRAs).\n- Inputs: natural language, tag sequences, JSON layout (layout is experimental). Prefer English (most training data); Chinese supported.\n\n### Limits / notes\n\n- Auto line breaks may not match intent ā€” use `\\n` or real newlines for precise text.\n- Turning off SLG looks more realistic but can collapse details; official negative-branch skip layer **22** (24/26 also help). kijai notes SLG can overblow anime styles.\n- Recommended negative prompt: `nsfw, worst quality, bad anatomy, watermark, jpeg artifact`\n- If VRAM is tight, lower resolution or use Sequential / CPU offload.\n- When training LoRA yourself, **freeze the text encoder**.\n- License: Apache 2.0.\n\n---\n\n## Sampling settings vs. official defaults\n\n| Setting | This template | OpenMOSS official |\n|---------|---------------|-------------------|\n| Steps | 25 | 50 |\n| CFG | 2.5 | 3 |\n| Sampler | er_sde | Euler |\n| Scheduler | simple | default |\n| Denoise | 1 | 1 |\n| Shift | 5 (model default) | 5 (same) |\n\nThis template follows the faster settings from kijai's ComfyUI MOSS-Image PR [#17095](https://github.com/comfyanonymous/ComfyUI/pull/17095). That is a speed trade-off, not a bug. For results closer to the official ones, try 50 steps and CFG 3.\n\n### Community notes\n\n- **Skip layer guidance:** can overblow anime styles; no alternative values were suggested (template keeps official layer 22). You can check the detail settings inside the subgraph\n- **Shift:** adjust with ModelSamplingAuraFlow or ModelSamplingFlux ā€” ModelSamplingSD3 uses a completely different scale. You can enter the subgraph to adjust it\n- **SFT banding:** experimental LoRA at about āˆ’0.5 strength: [moss_image_sft_lora_blocks_21_to_49ā€¦](https://huggingface.co/Kijai/MOSS-image_experimental/blob/main/loras/moss_image_sft_lora_blocks_21_to_49_rank_512_bf16.safetensors)\n" }, "color": "#222", "bgcolor": "#000" } ], "links": [ [ 22, 18, 0, 9, 0, "IMAGE" ], [ 40, 19, 0, 18, 3, "INT" ], [ 41, 19, 1, 18, 4, "INT" ] ], "groups": [], "definitions": { "subgraphs": [ { "id": "3301b744-2960-415a-83b0-fe8a89d99fdb", "version": 1, "state": { "lastGroupId": 5, "lastNodeId": 24, "lastLinkId": 52, "lastRerouteId": 0 }, "revision": 0, "config": {}, "name": "Text to Image (MOSS-Image)", "inputNode": { "id": -10, "bounding": [ -880, -120, 197.865234375, 328 ] }, "outputNode": { "id": -20, "bounding": [ 1630, -10, 128, 68 ] }, "inputs": [ { "id": "35437e31-565b-4d42-bc06-d95c5929e7d8", "name": "value", "type": "STRING", "linkIds": [ 25 ], "label": "positive_prompt", "pos": [ -706.134765625, -96 ] }, { "id": "853a4192-ebce-42de-bdff-77c36b59cc1a", "name": "switch", "type": "BOOLEAN", "linkIds": [ 26 ], "label": "enable_prompt_upsampler", "pos": [ -706.134765625, -76 ] }, { "id": "c11f1876-7b42-4679-a736-8deaa8f21fd7", "name": "text", "type": "STRING", "linkIds": [ 43 ], "label": "negative_prompt", "pos": [ -706.134765625, -56 ] }, { "id": "1514590b-7a78-450b-82b4-cfbe7820a9db", "name": "width", "type": "INT", "linkIds": [ 27 ], "pos": [ -706.134765625, -36 ] }, { "id": "f2e18bfe-65b0-4d24-9c78-bec0b9573e97", "name": "height", "type": "INT", "linkIds": [ 28 ], "pos": [ -706.134765625, -16 ] }, { "id": "32c26da2-7e60-44bf-851a-72690937b4b4", "name": "seed", "type": "INT", "linkIds": [ 30 ], "pos": [ -706.134765625, 4 ] }, { "id": "dae5d571-ced0-46d6-8a0a-5a71b8ca945a", "name": "steps", "type": "INT", "linkIds": [ 42 ], "pos": [ -706.134765625, 24 ] }, { "id": "6c064d6f-9925-4808-83cc-7e3409190ca7", "name": "cfg", "type": "FLOAT", "linkIds": [ 44 ], "pos": [ -706.134765625, 44 ] }, { "id": "46e71c72-9f95-4c32-89d2-2dab4b63b284", "name": "sampler_name", "type": "COMBO", "linkIds": [ 33 ], "pos": [ -706.134765625, 64 ] }, { "id": "8ba10fc2-8e73-4ece-8f77-27515bbe2698", "name": "scheduler", "type": "COMBO", "linkIds": [ 34 ], "pos": [ -706.134765625, 84 ] }, { "id": "71ffd624-a04e-403f-ae06-300f5896f34c", "name": "switch_1", "type": "BOOLEAN", "linkIds": [ 52 ], "label": "SLG", "pos": [ -706.134765625, 164 ] }, { "id": "bc8e2627-1b51-4333-9d88-0cb6307f7545", "name": "unet_name", "type": "COMBO", "linkIds": [ 35 ], "label": "unet_model", "pos": [ -706.134765625, 104 ] }, { "id": "fe422dee-acbe-4c34-898b-62bb878ddb83", "name": "clip_name", "type": "COMBO", "linkIds": [ 36 ], "label": "clip_model", "pos": [ -706.134765625, 124 ] }, { "id": "29a73fdf-7b55-4612-8966-bba0b4cab854", "name": "vae_name", "type": "COMBO", "linkIds": [ 37 ], "label": "vae_model", "pos": [ -706.134765625, 144 ] } ], "outputs": [ { "id": "fa9fefe4-677e-4f8b-aba5-35459b097c37", "name": "IMAGE", "type": "IMAGE", "linkIds": [ 6 ], "localized_name": "IMAGE", "pos": [ 1654, 14 ] } ], "widgets": [], "nodes": [ { "id": 17, "type": "PrimitiveStringMultiline", "pos": [ -140, -780 ], "size": [ 490, 280 ], "flags": {}, "order": 14, "mode": 0, "inputs": [ { "localized_name": "value", "name": "value", "type": "STRING", "widget": { "name": "value" }, "link": 25 } ], "outputs": [ { "localized_name": "STRING", "name": "STRING", "type": "STRING", "links": [ 15, 17 ] } ], "title": "Text (User Prompt)", "properties": { "Node name for S&R": "PrimitiveStringMultiline" }, "widgets_values": [ "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle" ], "widgets_values_named": { "value": "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle" } }, { "id": 14, "type": "PrimitiveStringMultiline", "pos": [ -140, -430 ], "size": [ 470, 270 ], "flags": { "collapsed": false }, "order": 0, "mode": 0, "inputs": [], "outputs": [ { "localized_name": "STRING", "name": "STRING", "type": "STRING", "links": [ 13 ] } ], "title": "Text (System prompt)", "properties": { "Node name for S&R": "PrimitiveStringMultiline" }, "widgets_values": [ "You turn a user's image idea into one detailed caption for a text-to-image model. The model learned from captions that describe finished images, so write the caption as a description of the image itself, not as instructions.\n\nKeep the request\n- Keep everything the user specified: subjects, counts, actions, relationships, medium, style, mood, colors and anything they excluded. Never contradict or drop it.\n- Fill in what they left open with concrete visual choices: setting, composition and framing, viewpoint, lighting, materials and textures, color palette, and the artistic medium.\n- Describe only what can be seen. Leave out sounds, smells, backstory and intentions.\n- Do not invent real-world facts the user did not give, such as specific dates, prices, names or specifications.\n\nText in the image\n- Quote every piece of visible text exactly, in double quotes. Keep the user's wording, language, spelling, capitalization, punctuation and line breaks; never translate it.\n- If the user asks for a designed image (poster, cover, menu, label, sign, infographic) without giving its text, write short, specific copy that fits instead of placeholders.\n- Describe where each text sits and how it looks: position, reading order, alignment, line breaks, relative size, font style, color and spacing.\n- Do not mention watermarks or artist signatures.\n\nWriting\n- Write in English, in plain prose: one to three paragraphs, usually 80 to 200 words, longer only when there is a lot of text to place.\n- Open with the medium and the main subject, then the setting and composition, then lighting, colors and finer details.\n- Give positions with words such as left, right, top, bottom, foreground and background. No coordinates, lists, headings, JSON or Markdown.\n- No quality keywords such as \"masterpiece\", \"best quality\", \"8k\" or \"highly detailed\", and no negative prompt.\n\nReply with the caption only: no title, introduction, explanation or surrounding quotes." ], "widgets_values_named": { "value": "You turn a user's image idea into one detailed caption for a text-to-image model. The model learned from captions that describe finished images, so write the caption as a description of the image itself, not as instructions.\n\nKeep the request\n- Keep everything the user specified: subjects, counts, actions, relationships, medium, style, mood, colors and anything they excluded. Never contradict or drop it.\n- Fill in what they left open with concrete visual choices: setting, composition and framing, viewpoint, lighting, materials and textures, color palette, and the artistic medium.\n- Describe only what can be seen. Leave out sounds, smells, backstory and intentions.\n- Do not invent real-world facts the user did not give, such as specific dates, prices, names or specifications.\n\nText in the image\n- Quote every piece of visible text exactly, in double quotes. Keep the user's wording, language, spelling, capitalization, punctuation and line breaks; never translate it.\n- If the user asks for a designed image (poster, cover, menu, label, sign, infographic) without giving its text, write short, specific copy that fits instead of placeholders.\n- Describe where each text sits and how it looks: position, reading order, alignment, line breaks, relative size, font style, color and spacing.\n- Do not mention watermarks or artist signatures.\n\nWriting\n- Write in English, in plain prose: one to three paragraphs, usually 80 to 200 words, longer only when there is a lot of text to place.\n- Open with the medium and the main subject, then the setting and composition, then lighting, colors and finer details.\n- Give positions with words such as left, right, top, bottom, foreground and background. No coordinates, lists, headings, JSON or Markdown.\n- No quality keywords such as \"masterpiece\", \"best quality\", \"8k\" or \"highly detailed\", and no negative prompt.\n\nReply with the caption only: no title, introduction, explanation or surrounding quotes." } }, { "id": 11, "type": "TextGenerate", "pos": [ 460, -780 ], "size": [ 400, 550 ], "flags": {}, "order": 10, "mode": 0, "inputs": [ { "localized_name": "clip", "name": "clip", "type": "CLIP", "link": 48 }, { "localized_name": "image", "name": "image", "shape": 7, "type": "IMAGE", "link": null }, { "localized_name": "video", "name": "video", "shape": 7, "type": "IMAGE", "link": null }, { "localized_name": "audio", "name": "audio", "shape": 7, "type": "AUDIO", "link": null }, { "localized_name": "system_prompt", "name": "system_prompt", "shape": 7, "type": "STRING", "link": 13 }, { "localized_name": "prompt", "name": "prompt", "type": "STRING", "widget": { "name": "prompt" }, "link": 17 } ], "outputs": [ { "localized_name": "generated_text", "name": "generated_text", "type": "STRING", "links": [ 19 ] }, { "localized_name": "thinking", "name": "thinking", "type": "STRING", "links": null } ], "properties": { "Node name for S&R": "TextGenerate" }, "widgets_values": [ "", 4096, "on", 0.7, 64, 0.95, 0.05, 1, 0, 0, false, true, "auto" ], "widgets_values_named": { "prompt": "", "max_length": 4096, "sampling_mode": "on", "sampling_mode.temperature": 0.7, "sampling_mode.top_k": 64, "sampling_mode.top_p": 0.95, "sampling_mode.min_p": 0.05, "sampling_mode.repetition_penalty": 1, "sampling_mode.seed": 0, "sampling_mode.presence_penalty": 0, "thinking": false, "use_default_template": true, "mtp": "auto" } }, { "id": 1, "type": "UNETLoader", "pos": [ -150, 10 ], "size": [ 350, 90 ], "flags": {}, "order": 2, "mode": 0, "inputs": [ { "localized_name": "unet_name", "name": "unet_name", "type": "COMBO", "widget": { "name": "unet_name" }, "link": 35 } ], "outputs": [ { "localized_name": "MODEL", "name": "MODEL", "type": "MODEL", "slot_index": 0, "links": [ 1, 51 ] } ], "properties": { "Node name for S&R": "UNETLoader", "models": [ { "name": "moss_image_sft_int8_convrot.safetensors", "url": "https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/diffusion_models/moss_image_sft_int8_convrot.safetensors", "directory": "diffusion_models" } ] }, "widgets_values": [ "moss_image_sft_int8_convrot.safetensors", "default" ], "widgets_values_named": { "unet_name": "moss_image_sft_int8_convrot.safetensors", "weight_dtype": "default" }, "color": "#223", "bgcolor": "#335" }, { "id": 4, "type": "CLIPTextEncode", "pos": [ 370, 30 ], "size": [ 450, 520 ], "flags": {}, "order": 5, "mode": 0, "inputs": [ { "localized_name": "clip", "name": "clip", "type": "CLIP", "link": 2 }, { "localized_name": "text", "name": "text", "type": "STRING", "widget": { "name": "text" }, "link": 20 } ], "outputs": [ { "localized_name": "CONDITIONING", "name": "CONDITIONING", "type": "CONDITIONING", "slot_index": 0, "links": [ 8 ] } ], "title": "Positive Prompt", "properties": { "Node name for S&R": "CLIPTextEncode" }, "widgets_values": [ "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle" ], "widgets_values_named": { "text": "Cinematic night street photograph: a model in an exaggerated balloon-sleeve white wool coat walking toward camera under a huge illuminated theatre marquee, the marquee bulbs spelling \"MOSS\" in tall capital letters, wet asphalt with long reflections, warm tungsten glow against cold blue night, anamorphic flare, 35mm film grain, low angle" }, "color": "#232", "bgcolor": "#353" }, { "id": 2, "type": "CLIPLoader", "pos": [ -150, 550 ], "size": [ 350, 140 ], "flags": {}, "order": 3, "mode": 0, "showAdvanced": false, "inputs": [ { "localized_name": "clip_name", "name": "clip_name", "type": "COMBO", "widget": { "name": "clip_name" }, "link": 36 } ], "outputs": [ { "localized_name": "CLIP", "name": "CLIP", "type": "CLIP", "slot_index": 0, "links": [ 2, 3, 48 ] } ], "properties": { "Node name for S&R": "CLIPLoader", "models": [ { "name": "qwen3.5_9b_w4a8_gptq.safetensors", "url": "https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/text_encoders/qwen3.5_9b_w4a8_gptq.safetensors", "directory": "text_encoders" } ] }, "widgets_values": [ "qwen3.5_9b_w4a8_gptq.safetensors", "moss_image", "default" ], "widgets_values_named": { "clip_name": "qwen3.5_9b_w4a8_gptq.safetensors", "type": "moss_image", "device": "default" }, "color": "#432", "bgcolor": "#653" }, { "id": 7, "type": "KSampler", "pos": [ 930, 0 ], "size": [ 290, 290 ], "flags": {}, "order": 8, "mode": 0, "inputs": [ { "localized_name": "model", "name": "model", "type": "MODEL", "link": 49 }, { "localized_name": "positive", "name": "positive", "type": "CONDITIONING", "link": 8 }, { "localized_name": "negative", "name": "negative", "type": "CONDITIONING", "link": 9 }, { "localized_name": "latent_image", "name": "latent_image", "type": "LATENT", "link": 39 }, { "localized_name": "seed", "name": "seed", "type": "INT", "widget": { "name": "seed" }, "link": 30 }, { "localized_name": "steps", "name": "steps", "type": "INT", "widget": { "name": "steps" }, "link": 42 }, { "localized_name": "cfg", "name": "cfg", "type": "FLOAT", "widget": { "name": "cfg" }, "link": 44 }, { "localized_name": "sampler_name", "name": "sampler_name", "type": "COMBO", "widget": { "name": "sampler_name" }, "link": 33 }, { "localized_name": "scheduler", "name": "scheduler", "type": "COMBO", "widget": { "name": "scheduler" }, "link": 34 } ], "outputs": [ { "localized_name": "LATENT", "name": "LATENT", "type": "LATENT", "slot_index": 0, "links": [ 4 ] } ], "properties": { "Node name for S&R": "KSampler" }, "widgets_values": [ 0, "randomize", 25, 2.5, "er_sde", "simple", 1 ], "widgets_values_named": { "seed": 0, "control_after_generate": "randomize", "steps": 25, "cfg": 2.5, "sampler_name": "er_sde", "scheduler": "simple", "denoise": 1 } }, { "id": 3, "type": "VAELoader", "pos": [ -150, 740 ], "size": [ 350, 60 ], "flags": {}, "order": 4, "mode": 0, "inputs": [ { "localized_name": "vae_name", "name": "vae_name", "type": "COMBO", "widget": { "name": "vae_name" }, "link": 37 } ], "outputs": [ { "localized_name": "VAE", "name": "VAE", "type": "VAE", "slot_index": 0, "links": [ 5 ] } ], "properties": { "Node name for S&R": "VAELoader", "models": [ { "name": "flux2-vae_small.safetensors", "url": "https://huggingface.co/Comfy-Org/MOSS-Image/resolve/main/vae/flux2-vae_small.safetensors", "directory": "vae" } ] }, "widgets_values": [ "flux2-vae_small.safetensors" ], "widgets_values_named": { "vae_name": "flux2-vae_small.safetensors" }, "color": "#323", "bgcolor": "#535" }, { "id": 13, "type": "PreviewAny", "pos": [ 1260, -780 ], "size": [ 350, 600 ], "flags": {}, "order": 11, "mode": 0, "inputs": [ { "localized_name": "source", "name": "source", "type": "*", "link": 21 } ], "outputs": [ { "localized_name": "STRING", "name": "STRING", "type": "STRING", "links": [ 20 ] } ], "title": "Preview as Text (final prompt)", "properties": { "Node name for S&R": "PreviewAny" }, "widgets_values": [], "widgets_values_named": {} }, { "id": 16, "type": "ComfySwitchNode", "pos": [ 920, -780 ], "size": [ 280, 110 ], "flags": {}, "order": 13, "mode": 0, "inputs": [ { "localized_name": "on_false", "name": "on_false", "shape": 7, "type": "STRING", "link": 15 }, { "localized_name": "on_true", "name": "on_true", "shape": 7, "type": "STRING", "link": 19 }, { "localized_name": "switch", "name": "switch", "type": "BOOLEAN", "widget": { "name": "switch" }, "link": 26 } ], "outputs": [ { "localized_name": "output", "name": "output", "type": "STRING", "links": [ 21 ] } ], "properties": { "Node name for S&R": "ComfySwitchNode" }, "widgets_values": [ true ], "widgets_values_named": { "switch": true } }, { "id": 6, "type": "EmptyFlux2LatentImage", "pos": [ -130, 960 ], "size": [ 300, 120 ], "flags": {}, "order": 7, "mode": 0, "inputs": [ { "localized_name": "width", "name": "width", "type": "INT", "widget": { "name": "width" }, "link": 27 }, { "localized_name": "height", "name": "height", "type": "INT", "widget": { "name": "height" }, "link": 28 } ], "outputs": [ { "localized_name": "LATENT", "name": "LATENT", "type": "LATENT", "slot_index": 0, "links": [ 39 ] } ], "properties": { "Node name for S&R": "EmptyFlux2LatentImage" }, "widgets_values": [ 1024, 1024, 1 ], "widgets_values_named": { "width": 1024, "height": 1024, "batch_size": 1 }, "color": "#323", "bgcolor": "#535" }, { "id": 5, "type": "CLIPTextEncode", "pos": [ 370, 610 ], "size": [ 460, 180 ], "flags": {}, "order": 6, "mode": 0, "inputs": [ { "localized_name": "clip", "name": "clip", "type": "CLIP", "link": 3 }, { "localized_name": "text", "name": "text", "type": "STRING", "widget": { "name": "text" }, "link": 43 } ], "outputs": [ { "localized_name": "CONDITIONING", "name": "CONDITIONING", "type": "CONDITIONING", "slot_index": 0, "links": [ 9 ] } ], "title": "Negative Prompt (empty = reference default)", "properties": { "Node name for S&R": "CLIPTextEncode" }, "widgets_values": [ "" ], "widgets_values_named": { "text": "" }, "color": "#223", "bgcolor": "#335" }, { "id": 8, "type": "VAEDecode", "pos": [ 1270, 0 ], "size": [ 230, 70 ], "flags": {}, "order": 9, "mode": 0, "inputs": [ { "localized_name": "samples", "name": "samples", "type": "LATENT", "link": 4 }, { "localized_name": "vae", "name": "vae", "type": "VAE", "link": 5 } ], "outputs": [ { "localized_name": "IMAGE", "name": "IMAGE", "type": "IMAGE", "slot_index": 0, "links": [ 6 ] } ], "properties": { "Node name for S&R": "VAEDecode" } }, { "id": 15, "type": "SkipLayerGuidanceDiTSimple", "pos": [ -150, 280 ], "size": [ 350, 220 ], "flags": {}, "order": 12, "mode": 0, "showAdvanced": true, "inputs": [ { "localized_name": "model", "name": "model", "type": "MODEL", "link": 1 } ], "outputs": [ { "localized_name": "MODEL", "name": "MODEL", "type": "MODEL", "links": [ 50 ] } ], "properties": { "Node name for S&R": "SkipLayerGuidanceDiTSimple" }, "widgets_values": [ "", "22", 0, 0.5 ], "widgets_values_named": { "double_layers": "", "single_layers": "22", "start_percent": 0, "end_percent": 0.5 } }, { "id": 24, "type": "ComfySwitchNode", "pos": [ -110, 140 ], "size": [ 270, 105.99999028338053 ], "flags": {}, "order": 1, "mode": 0, "inputs": [ { "localized_name": "on_false", "name": "on_false", "shape": 7, "type": "MODEL", "link": 51 }, { "localized_name": "on_true", "name": "on_true", "shape": 7, "type": "MODEL", "link": 50 }, { "localized_name": "switch", "name": "switch", "type": "BOOLEAN", "widget": { "name": "switch" }, "link": 52 } ], "outputs": [ { "localized_name": "output", "name": "output", "type": "MODEL", "links": [ 49 ] } ], "properties": { "Node name for S&R": "ComfySwitchNode" }, "widgets_values": [ false ], "widgets_values_named": { "switch": false } } ], "groups": [ { "id": 1, "title": "Models", "bounding": [ -210, -60, 460, 880 ], "color": "#3f789e", "flags": {} }, { "id": 2, "title": "Prompt", "bounding": [ 290, -60, 580, 880 ], "color": "#3f789e", "flags": {} }, { "id": 3, "title": "Image Size", "bounding": [ -210, 850, 460, 300 ], "color": "#3f789e", "flags": {} }, { "id": 4, "title": "Prompt Enhancer", "bounding": [ -220, -880, 1900, 840 ], "color": "#444", "flags": {} } ], "links": [ { "id": 2, "origin_id": 2, "origin_slot": 0, "target_id": 4, "target_slot": 0, "type": "CLIP" }, { "id": 20, "origin_id": 13, "origin_slot": 0, "target_id": 4, "target_slot": 1, "type": "STRING" }, { "id": 1, "origin_id": 1, "origin_slot": 0, "target_id": 15, "target_slot": 0, "type": "MODEL" }, { "id": 8, "origin_id": 4, "origin_slot": 0, "target_id": 7, "target_slot": 1, "type": "CONDITIONING" }, { "id": 9, "origin_id": 5, "origin_slot": 0, "target_id": 7, "target_slot": 2, "type": "CONDITIONING" }, { "id": 21, "origin_id": 16, "origin_slot": 0, "target_id": 13, "target_slot": 0, "type": "STRING" }, { "id": 15, "origin_id": 17, "origin_slot": 0, "target_id": 16, "target_slot": 0, "type": "STRING" }, { "id": 19, "origin_id": 11, "origin_slot": 0, "target_id": 16, "target_slot": 1, "type": "STRING" }, { "id": 3, "origin_id": 2, "origin_slot": 0, "target_id": 5, "target_slot": 0, "type": "CLIP" }, { "id": 4, "origin_id": 7, "origin_slot": 0, "target_id": 8, "target_slot": 0, "type": "LATENT" }, { "id": 5, "origin_id": 3, "origin_slot": 0, "target_id": 8, "target_slot": 1, "type": "VAE" }, { "id": 6, "origin_id": 8, "origin_slot": 0, "target_id": -20, "target_slot": 0, "type": "IMAGE" }, { "id": 39, "origin_id": 6, "origin_slot": 0, "target_id": 7, "target_slot": 3, "type": "LATENT" }, { "id": 48, "origin_id": 2, "origin_slot": 0, "target_id": 11, "target_slot": 0, "type": "CLIP" }, { "id": 13, "origin_id": 14, "origin_slot": 0, "target_id": 11, "target_slot": 4, "type": "STRING" }, { "id": 17, "origin_id": 17, "origin_slot": 0, "target_id": 11, "target_slot": 5, "type": "STRING" }, { "id": 49, "origin_id": 24, "origin_slot": 0, "target_id": 7, "target_slot": 0, "type": "MODEL" }, { "id": 50, "origin_id": 15, "origin_slot": 0, "target_id": 24, "target_slot": 1, "type": "MODEL" }, { "id": 51, "origin_id": 1, "origin_slot": 0, "target_id": 24, "target_slot": 0, "type": "MODEL" }, { "id": 25, "origin_id": -10, "origin_slot": 0, "target_id": 17, "target_slot": 0, "type": "STRING" }, { "id": 26, "origin_id": -10, "origin_slot": 1, "target_id": 16, "target_slot": 2, "type": "BOOLEAN" }, { "id": 43, "origin_id": -10, "origin_slot": 2, "target_id": 5, "target_slot": 1, "type": "STRING" }, { "id": 27, "origin_id": -10, "origin_slot": 3, "target_id": 6, "target_slot": 0, "type": "INT" }, { "id": 28, "origin_id": -10, "origin_slot": 4, "target_id": 6, "target_slot": 1, "type": "INT" }, { "id": 30, "origin_id": -10, "origin_slot": 5, "target_id": 7, "target_slot": 4, "type": "INT" }, { "id": 42, "origin_id": -10, "origin_slot": 6, "target_id": 7, "target_slot": 5, "type": "INT" }, { "id": 44, "origin_id": -10, "origin_slot": 7, "target_id": 7, "target_slot": 6, "type": "FLOAT" }, { "id": 33, "origin_id": -10, "origin_slot": 8, "target_id": 7, "target_slot": 7, "type": "COMBO" }, { "id": 34, "origin_id": -10, "origin_slot": 9, "target_id": 7, "target_slot": 8, "type": "COMBO" }, { "id": 52, "origin_id": -10, "origin_slot": 10, "target_id": 24, "target_slot": 2, "type": "BOOLEAN" }, { "id": 35, "origin_id": -10, "origin_slot": 11, "target_id": 1, "target_slot": 0, "type": "COMBO" }, { "id": 36, "origin_id": -10, "origin_slot": 12, "target_id": 2, "target_slot": 0, "type": "COMBO" }, { "id": 37, "origin_id": -10, "origin_slot": 13, "target_id": 3, "target_slot": 0, "type": "COMBO" } ], "extra": {} } ] }, "config": {}, "extra": { "frontendVersion": "1.55.17", "ds": { "scale": 0.5123746044423974, "offset": [ 1458.7144197108835, 566.1603684330503 ] } }, "version": 0.4 }