💡Steps to Follow:
Step 1: Photo Collection
• Gather 15+ photos of the couple
• Include different: poses, outfits, and locations
• Name format: "a_photo_of_(name)"
• Save all photos in high quality
• TIP: Ensure clear face visibility
Step 2: Replicate Setup
• Go to replicate and search: lucataco/ai-toolkit
• Select 'Train' tab and create new model
• Name your project and upload photos as ZIP file
• Set steps to 1200
Step 3: Hugging Face Setup
• Visit huggingface.co and login
• Go to Settings
• Create new access token and copy token
• Create new model and set permissions to public
• Copy model path
Step 4: Training Proces
• Paste HF token in Replicate
• Add model path to repo_id and click 'Train'
• Wait ~30 minutes and check training progress
• TIP: Don't close browser
Step 5: Prompt Engineering with ChatGPT
• Ask ChatGPT for creative ideas:
- Unique locations (beaches, mountains, cities)
- Cultural attire combinations
- Romantic poses and expressions
- Time of day variations
• IMPORTANT: Always start prompts with "a photo of (saved file name)"
• Example prompt for ChatGPT:
"Generate 5 detailed prompts for pre-wedding shoots including location, attire, lighting, and mood"
Step 6: Generation
• Search flux-dev-lora on Replicate
• Paste HF model path
• Use ChatGPT-generated prompts
• Generate images and download results
You can check my Instagram video here for more details.
Комментариев нет:
Отправить комментарий
Примечание. Отправлять комментарии могут только участники этого блога.