Skip to content

QUICKSTART.md

Latest commit

 

History

History
179 lines (141 loc) · 4.9 KB

QUICKSTART.md

File metadata and controls

179 lines (141 loc) · 4.9 KB
Raw

Quick Start Guide

Time to complete: 4-8 hours for basic setup + testing

🚀 Fastest Path to Working Demo

Step 1: Setup RunPod (1 hour)

1. Create RunPod account → Add $45 credits
2. Deploy RTX 4090 pod with Pytorch template
3. SSH into pod
4. Run setup script (see below)

Quick Setup Script (paste into RunPod terminal):

#!/bin/bash
cd /workspace
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt xformers

# Install InstantID
cd custom_nodes
git clone https://github.com/cubiq/ComfyUI_InstantID.git
cd ComfyUI_InstantID && pip install -r requirements.txt && cd ..
cd ..

# Download models
mkdir -p models/instantid models/insightface/models
cd models/checkpoints
wget https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors
cd ../instantid
wget https://huggingface.co/InstantX/InstantID/resolve/main/ip-adapter.bin
wget https://huggingface.co/InstantX/InstantID/resolve/main/ControlNetModel/diffusion_pytorch_model.safetensors -O controlnet.safetensors
cd ../insightface/models
wget https://huggingface.co/MonsterMMORPG/tools/resolve/main/antelopev2.zip && unzip antelopev2.zip
cd /workspace/ComfyUI

# Start ComfyUI
python main.py --listen 0.0.0.0 --port 8188

Step 2: Load ComfyUI Workflow (5 minutes) ✅

GOOD NEWS: Workflow đã sẵn sàng!

  1. Access ComfyUI at your RunPod URL
  2. Download workflows/workflow_api.json
  3. In ComfyUI: Click Load → Select workflow_api.json
  4. Upload reference image to Node 1 (LoadImage)
  5. Click Queue Prompt to test
  6. Done! ✅

What's included:

  • ✅ Pony Diffusion XL v6 (best for NSFW)
  • ✅ SDXL VAE (separate, better quality)
  • ✅ InstantID (ip_weight: 0.75, cn_strength: 0.5)
  • ✅ Optimized settings (steps: 30, cfg: 7.0)
  • ✅ Score prompts for Pony
  • ✅ 1.5x upscaling

Step 3: Setup Local Environment (30 min)

# Clone this repo (or your fork)
git clone https://github.com/yourusername/nectar.git
cd nectar

# Backend
cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate
pip install -r requirements.txt
cp .env.example .env
# Edit .env: Add your ANTHROPIC_API_KEY and COMFYUI_URL

# Frontend
cd ../frontend
npm install
cp .env.local.example .env.local

Step 4: Run Application (5 min)

Terminal 1 (Backend):

cd backend
source venv/bin/activate
python app/main.py

Terminal 2 (Frontend):

cd frontend
npm run dev

Open http://localhost:3000

Step 5: Generate Samples (2-4 hours)

  1. Upload 2 reference faces (1 male, 1 female)
  2. Generate 5+ scenes per character:
    • Coffee shop
    • Beach
    • Gym
    • Bedroom
    • Outdoor
  3. Generate 2-3 NSFW examples per character
  4. Save all to generated_images/

Step 6: Record Demo & Document (1 hour)

  1. Record 3-5 minute demo video showing:
    • Setup process
    • Natural conversation triggering images
    • Consistency comparison
  2. Update README with your technical decisions
  3. Create comparison grids

Step 7: Submit

git add .
git commit -m "Complete Nectar AI assessment"
git push origin main
# Share repo with 0xtaozi and 0xmihutao

📋 Minimal Checklist

  • RunPod setup complete (ComfyUI running)
  • Workflow built and tested
  • Local app running (backend + frontend)
  • 5+ images per character (male + female)
  • 2-3 NSFW examples per character
  • Demo video recorded
  • README updated with technical justifications
  • GitHub repo shared

🔥 Pro Tips

  1. Save Time: Workflow đã ready - chỉ cần load và test!
  2. Face Consistency: Current settings (ip_weight: 0.75) đã optimized
  3. NSFW Quality: Pony XL v6 + 30-35 steps = excellent anatomy
  4. Generation Speed: Keep models in VRAM, don't reload
  5. Cost: Stop RunPod when not using (~$15-20 total expected)
  6. VAE Quality: Using separate SDXL VAE for better colors

🆘 Quick Troubleshooting

Weird colors/artifacts: Make sure sdxl_vae.safetensors is loaded (Node 21) Face not consistent: Increase ip_weight from 0.75 to 0.80-0.85 Abstract/noise output: Check VAE file exists, try lowering cfg to 6.5 Out of memory: Reduce resolution to 768x768 Slow generation: Already optimized at 30 steps Model not found: Check Pony XL filename is exact: ponyDiffusionV6XL_v6StartWithThisOne.safetensors

See workflows/TROUBLESHOOTING.md for detailed fixes!

📚 Full Documentation

⏱️ Time Breakdown

  • RunPod + ComfyUI setup: 1-2 hours
  • Workflow building: 1 hour
  • Local setup: 30 min
  • Sample generation: 2-4 hours
  • Demo + docs: 1 hour
  • Total: 6-9 hours

Start now and you can complete this in 1-2 focused work sessions! 🚀