This example shows how to create and use two personas together.
# Create Alice
python scripts/persona_manager.py add --name "Alice Chen" \
--description "Professional woman, Asian, 28 years old"
# Create Bob
python scripts/persona_manager.py add --name "Bob Smith" \
--description "Business executive, Caucasian, 35 years old"This creates:
persona-alice_chenwith trigger wordpersona-alice_chenpersona-bob_smithwith trigger wordpersona-bob_smith
Place 20-30 high-quality images in:
training_data/persona-alice_chen/raw/training_data/persona-bob_smith/raw/
# Process Alice's images
python scripts/prepare_training_data.py --persona-id persona-alice_chen
# Train Alice
./scripts/train_lora.sh persona-alice_chen
# Process Bob's images
python scripts/prepare_training_data.py --persona-id persona-bob_smith
# Train Bob
./scripts/train_lora.sh persona-bob_smith# Create workflow for Alice
python persona_gen.py create-workflow --persona-id persona-alice_chen --type image
# Start ComfyUI
./scripts/run_comfyui.sh
# Load the workflow and use prompts like:
"professional headshot of persona-alice_chen, business attire, studio lighting"# Generate multi-persona workflow
python scripts/persona_manager.py generate-workflow \
persona-alice_chen persona-bob_smith \
--prompt "business meeting" \
--output workflows/alice_bob_meeting.json
# Load in ComfyUI and use prompts like:
"persona-alice_chen and persona-bob_smith shaking hands in modern office,
professional attire, natural lighting, depth of field""cinematic photo, persona-alice_chen presenting to persona-bob_smith,
conference room, projector screen in background, business formal attire,
professional lighting, 8k, ultra detailed"
"candid photo, persona-alice_chen and persona-bob_smith having coffee,
casual business attire, modern cafe, warm lighting, bokeh background"
"professional group photo, persona-alice_chen on left wearing navy suit,
persona-bob_smith on right wearing gray suit, corporate lobby background,
both smiling, studio lighting"
- Consistent Training: Use similar lighting and quality for all training images
- Unique Features: Include diverse angles and expressions in training data
- Clear Descriptions: Be specific about positioning and clothing
- Test Individually: Verify each persona works alone before combining
- Adjust Strengths: Lower LoRA strengths (0.6-0.7) when using multiple personas