comfyui_siliconflow

🎨 ComfyUI SiliconFlow Specialized Nodes

A production-grade collection of specialized ComfyUI nodes for the SiliconFlow Image Generation API.

This extension provides granular nodes for every major model variant in the SiliconFlow documentation. Each node is surgically tailored to expose every single parameter supported by that specific model, ensuring full control without UI clutter.

---

🚀 Key Features

• Exhaustive Control: Every mandatory and optional parameter from the SiliconFlow API is exposed.
• Zero UI Clutter: Parameters irrelevant to a model are physically hidden via specialized nodes.
• Live Tooltips: Hover over any parameter in ComfyUI for real-time documentation.
• Metadata Recovery: All nodes output the actual_seed used by the API for perfect reproducibility.
• Native Batching: Returns standard ComfyUI IMAGE batch tensors.

---

🧩 Comprehensive Node & Parameter Reference

All nodes return (IMAGE, INT) representing the image tensor and the actual_seed.

1. FLUX.2 Series

🎨 SiliconFlow — FLUX.2 Pro

High-quality professional generation.

• model (Required): black-forest-labs/FLUX.2-pro
• prompt (Required): Text description of the image.
• image_size (Required): [512x512, 768x1024, 1024x768, 576x1024, 1024x576]
• seed (Required): 0 - 9999999999.
• output_format (Required): [png, jpeg]

🎨 SiliconFlow — FLUX.2 Flex

Flexible generation with CFG and step control.

• model (Required): black-forest-labs/FLUX.2-flex
• prompt (Required): Text description.
• image_size (Required): [512x512, 768x1024, 1024x768, 576x1024, 1024x576]
• seed (Required): 0 - 9999999999.
• num_inference_steps (Optional): 1 - 50 (Default: 25).
• cfg (Optional): 0.1 - 20.0 (Default: 4.0).
• output_format (Optional): [png, jpeg]

---

2. FLUX.1 Series (Advanced)

🎨 SiliconFlow — FLUX-1.1 Pro

High-performance resolution control.

• model (Required): black-forest-labs/FLUX-1.1-pro
• prompt (Required): Text description.
• width (Required): 256 - 1440 (Must be multiple of 32, Default: 1024).
• height (Required): 256 - 1440 (Must be multiple of 32, Default: 768).
• seed (Required): 0 - 9999999999.
• image_prompt (Optional): IMAGE input for visual guidance.
• prompt_upsampling (Optional): True/False (Automatic creative expansion).
• safety_tolerance (Optional): 0 - 6 (0: Strictest, 6: Lenient, Default: 2).
• output_format (Optional): [png, jpeg]

🎨 SiliconFlow — FLUX-1.1 Pro Ultra

The flagship generation node with every available feature.

• model (Required): black-forest-labs/FLUX-1.1-pro-Ultra
• prompt (Required): Text description.
• image_size (Required): [1024x1024, 960x1280, 768x1024, 720x1440, 720x1280, others]
• seed (Required): 0 - 9999999999.
• negative_prompt (Optional): Elements to avoid in the image.
• batch_size (Optional): 1 - 4 (Generate multiple images in one go).
• aspect_ratio (Optional): String between 21:9 and 9:21 (Default: 1:1).
• safety_tolerance (Optional): 0 - 6 (Default: 2).
• output_format (Optional): [png, jpeg]
• raw (Optional): True/False (Generates more natural, photographic textures).
• image_prompt (Optional): IMAGE to remix.
• image_prompt_strength (Optional): 0.0 - 1.0 (Blend between text and image prompt, Default: 0.1).

🎨 SiliconFlow — FLUX.1 Dev

• model (Required): black-forest-labs/FLUX.1-dev
• prompt (Required): Text description.
• image_size (Required): [1024x1024, 960x1280, 768x1024, 720x1440, 720x1280, others]
• num_inference_steps (Required): 1 - 30 (Default: 20).
• seed (Required): 0 - 9999999999.
• prompt_enhancement (Optional): True/False (Optimize for model-friendliness).

🎨 SiliconFlow — FLUX.1 Schnell

• model (Required): black-forest-labs/FLUX.1-schnell
• prompt (Required): Text description.
• image_size (Required): [1024x1024, 512x1024, 768x512, 768x1024, 1024x576, 576x1024]
• seed (Required): 0 - 9999999999.
• prompt_enhancement (Optional): True/False.

---

3. Kontext (Image-to-Image)

🎨 SiliconFlow — FLUX.1 Kontext

Advanced context-aware generation.

• model (Required): black-forest-labs/FLUX.1-Kontext-max, black-forest-labs/FLUX.1-Kontext-pro
• prompt (Required): Text description.
• image (Required): IMAGE input (The base image).
• seed (Required): 0 - 9999999999.
• aspect_ratio (Optional): String between 21:9 and 9:21 (Default: 1:1).
• output_format (Optional): [png, jpeg]
• prompt_upsampling (Optional): True/False.
• safety_tolerance (Optional): 0 - 6 (Note: Img2Img is usually capped at 2).

🎨 SiliconFlow — FLUX.1 Kontext Dev

• model (Required): black-forest-labs/FLUX.1-Kontext-dev
• prompt (Required): Text description.
• image (Required): IMAGE input.
• seed (Required): 0 - 9999999999.
• prompt_enhancement (Optional): True/False.

---

4. Qwen & Z-Image

🎨 SiliconFlow — Qwen Image

Complete Qwen implementation supporting both generation and editing.

• model (Required): Qwen/Qwen-Image, Qwen/Qwen-Image-Edit
• prompt (Required): Text description.
• image_size (Required): [1328x1328, 1664x928, 928x1664, 1472x1140, 1140x1472, 1584x1056, 1056x1584]
• seed (Required): 0 - 9999999999.
• negative_prompt (Optional): Elements to exclude.
• batch_size (Optional): 1 - 4.
• num_inference_steps (Optional): 1 - 100 (Default: 20).
• guidance_scale (Optional): 0.0 - 20.0 (Stricter vs Creative adherence, Default: 7.5).
• cfg (Optional): 0.1 - 20.0 (Required for text generation, Default: 4.0).
• image (Optional): IMAGE input for editing models.

🎨 SiliconFlow — Z-Image

Ultra-fast turbo generation.

• model (Required): Tongyi-MAI/Z-Image-Turbo
• prompt (Required): Text description.
• image_size (Required): [512x512, 768x1024, 1024x576, 576x1024]
• seed (Required): 0 - 9999999999.
• negative_prompt (Optional): Elements to avoid.

---

🛠️ Setup

1. API Key: Create apikey.txt in the root folder and paste your key.
2. Category: All nodes are under SiliconFlow in the ComfyUI menu.

⚖️ License

MIT