ComfyUI Extension: Neural Atelier
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.
Professional ComfyUI node pack for AI-powered fashion visualization. Transform flat sketches into product photos, ghost mannequin views, editorial shots, e-commerce images and color variants using Gemini AI.
Looking for a different extension?
Custom Nodes (5)
README
Neural Atelier - ComfyUI Custom Node Pack
Overview
Neural Atelier NA01 is a professional ComfyUI node for fashion visualization and product photography. It orchestrates Google's Gemini AI models to transform sketches, materials, and briefs into photorealistic product images.
Architecture
User Input → Gemini 3 Flash (Prompt Orchestration) → Nano Banana Pro (Image Generation) → Output
- Gemini 3 Flash (
gemini-3-flash-preview): Analyzes your inputs and generates an optimized prompt - Nano Banana Pro (
gemini-3-pro-image-preview): Generates the final high-quality image
Installation
-
Copy the entire
comfyui_neural_atelierfolder to your ComfyUI custom nodes directory:ComfyUI/custom_nodes/comfyui_neural_atelier/ -
Install dependencies:
pip install -r requirements.txt -
Restart ComfyUI
Nodes
NA01 Sketch to Photo Orchestrator
The main node that handles the complete workflow.
Inputs:
| Input | Type | Required | Description | |-------|------|----------|-------------| | prompt_pack | Dropdown | Yes | Select a prompt pack from configs/ | | brief_text | Text | Yes | Your creative brief | | gemini_api_key | String | Yes | Your Gemini API key | | api_key_status | Display | Yes | Shows API key verification status | | aspect_ratio | Dropdown | Yes | 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9 | | resolution | Dropdown | Yes | 1K, 2K, 4K (image quality) | | top_p | Float | Yes | Creativity parameter (0.0-1.0) | | talent_image | Image | No | Reference talent/model image | | flat_sketch_image | Image | No | Technical sketch | | material_image | Image | No | Material/texture reference | | pattern_image | Image | No | Pattern reference | | template_master_1 | Image | No | Template image 1 | | template_master_2 | Image | No | Template image 2 |
Outputs:
image: Generated imagelog: Execution summary with run_id, timings, and file paths
NA01 Verify API Key
Utility node to verify your Gemini API key.
NA01 Get Profiles
Utility node to list available profiles for a prompt pack.
Prompt Packs
Located in configs/, organized by use case:
- 01_Senior_Technical_Fashion_Designer: Sketch-to-photo translation
- 02_Physics_Simulation: Ghost mannequin and hanger presentations
- 03_Macro_Details: 2x2 detail grid compositions
- 04_Technical_Illustration: Isometric schematic diagrams
- 05_Exploded_View: Photorealistic exploded view infographics
Output Logging
All executions are logged to:
ComfyUI/output/neural_atelier/NA01/YYYY-MM-DD/[run_id]/
Files per run:
run.json: Complete execution metadatanano_banana_prompt.txt: Generated promptgemini_flash_response_raw.txt: Raw API responserequest_manifest.json: Input configuration
API Key Configuration
You can provide your API key in two ways:
- Node Input: Enter directly in the
gemini_api_keyfield - Environment Variable: Set
GEMINI_API_KEYin your environment
Security
- API keys are never written to disk
- All logs automatically redact authentication data
- Keys can be verified without exposing them
Requirements
- ComfyUI (latest version recommended)
- Python 3.10+
- google-genai >= 1.0.0
- Pillow >= 9.0.0
- torch >= 2.0.0
Version
1.0.0
License
MIT License
Where to Find All Workflows
If you want access to the full workflow collection, the latest updates, exclusive releases, and ongoing development for Neural Atelier, you can find everything on my Patreon:
Join here: https://www.patreon.com/c/SergioValsecchi
Patreon is the main hub for all workflow packs, custom node updates, advanced R&D, and new tools as they are released.
Run ComfyUI workflows without the setup
No installs, no CUDA version roulette, no GPU sitting idle on your bill. Bring a workflow and run it in the browser.