Usage Examples - jmpijll/discomfy GitHub Wiki

Usage Examples

Practical examples for using DisComfy commands and features.

🎨 Basic Image Generation

Simple Generation

/generate prompt:a beautiful sunset over the ocean

What happens:

  1. Interactive setup view appears
  2. Choose model (Flux, Flux Krea, or HiDream)
  3. Optionally select a LoRA
  4. Click "Generate"
  5. Watch real-time progress
  6. Image appears with action buttons

Generation with Parameters

/generate prompt:cyberpunk cityscape at night width:1024 height:768 steps:30 cfg:7.5 batch_size:2

Parameters explained:

  • width:1024 - Image width in pixels
  • height:768 - Image height in pixels
  • steps:30 - 30 sampling steps for quality
  • cfg:7.5 - Guidance scale
  • batch_size:2 - Generate 2 variations

Using LoRAs

/generate prompt:anime character in magical forest lora:anime_style_v2 lora_strength:0.8

LoRA features:

  • lora:anime_style_v2 - Name of LoRA to use
  • lora_strength:0.8 - 80% strength (0.0-2.0 range)

✏️ Image Editing

High-Quality Editing (Flux Kontext)

For detailed, high-quality edits:

/editflux image:<upload_photo> prompt:add sunglasses and a hat steps:25

Best for:

  • Detailed edits
  • Complex changes
  • Final quality edits
  • When you have time (1-3 minutes)

Tips:

  • Use 20-30 steps for best quality
  • Be specific in prompts
  • More steps = better detail

Ultra-Fast Editing (Qwen 2.5 VL)

For quick iterations:

/editqwen image:<upload_photo> prompt:change the sky to sunset steps:8

Best for:

  • Rapid iteration
  • Testing ideas
  • Quick changes
  • When speed matters (30-60 seconds)

Tips:

  • Use 4-8 steps for fastest results
  • 8-12 steps for better quality
  • Perfect for experimentation

Editing Generated Images

Post-generation editing:

  1. Generate an image with /generate
  2. Click ✏️ Flux Edit button for high quality
  3. OR click ⚡ Qwen Edit button for speed
  4. Enter edit prompt in modal
  5. Adjust steps if needed
  6. Submit and watch progress

Example prompts:

  • "make the character smile"
  • "add a dramatic storm in the background"
  • "change hair color to blue"
  • "add magical sparkles around the subject"
  • "make it nighttime instead of day"

🔍 Image Upscaling

Basic Upscaling

From generated image:

  1. Generate or upload image
  2. Click 🔍 Upscale button
  3. Modal appears with options:
    • Upscale Ratio: Choose 2x, 4x, or 8x
    • Denoise Strength: 0.5 (default)
    • Sampling Steps: 20 (default)
  4. Click Submit

Upscaling Scenarios

Quick 2x Upscale:

  • Ratio: 2x
  • Denoise: 0.3
  • Steps: 15
  • Use: Fast upscale, minimal changes

Standard 4x Upscale:

  • Ratio: 4x
  • Denoise: 0.5
  • Steps: 20
  • Use: Balanced quality and speed

Maximum 8x Detail Enhancement:

  • Ratio: 8x
  • Denoise: 0.7
  • Steps: 30
  • Use: Maximum detail, poster quality

Denoise Strength Guide:

  • 0.1-0.3 - Minimal change, preserve original
  • 0.4-0.6 - Balanced enhancement
  • 0.7-1.0 - Strong detail enhancement

🎬 Video Generation

Basic Animation

From generated image:

  1. Generate an image with /generate
  2. Click 🎬 Animate button
  3. Modal shows options:
    • Frame Count: 121 frames (default)
    • Strength: 0.7
    • Steps: 25
    • Animation Prompt: (original prompt pre-filled)
  4. Click Submit
  5. Wait 2-10 minutes depending on settings

Custom Animation Prompts

Change how the image animates:

Slow Pan:

Animation Prompt: slowly pan across the scene from left to right

Zoom Effect:

Animation Prompt: camera slowly zooms in on the subject

Rotation:

Animation Prompt: gentle 360 degree rotation around the subject

Dynamic Movement:

Animation Prompt: dynamic movement with parallax effect and depth

Dramatic Effect:

Animation Prompt: dramatic camera movement with zoom and rotation

Frame Count Selection

81 frames (~2 seconds):

  • Fastest generation
  • Short loops
  • Quick previews

121 frames (~3 seconds):

  • Standard animation
  • Good balance
  • Most common choice

161 frames (~5 seconds):

  • Longer animation
  • More dramatic
  • Better for complex motion
  • Takes longer to generate

🎯 Advanced Generation

Seed Control

Reproduce a good result:

  1. Generate image with random seed
  2. Note the seed from image info
  3. Reuse seed for similar results:
/generate prompt:your prompt seed:123456789 steps:30

Using seeds:

  • -1 = Random seed (default)
  • Same seed + same params = Similar result
  • Same seed + different prompt = Similar composition

Batch Generation

Generate multiple variations:

/generate prompt:fantasy castle on a hill batch_size:4

What happens:

  • Generates 4 different images
  • All with same parameters
  • Different seeds automatically
  • Posted as multiple messages

Best practices:

  • Use batch_size 2-4 for variations
  • More images = longer wait
  • Good for exploring ideas

CFG Scale Experiments

Low CFG (3-5):

/generate prompt:abstract art painting cfg:4
  • More creative interpretation
  • Less adherence to prompt
  • More artistic freedom

Medium CFG (6-8):

/generate prompt:realistic portrait cfg:7.5
  • Balanced results
  • Good prompt adherence
  • Recommended default

High CFG (9-15):

/generate prompt:photorealistic cityscape cfg:12
  • Very strict prompt following
  • Can be over-saturated
  • Use carefully

🔄 Complete Workflow Examples

Workflow 1: Character Creation

  1. Initial Generation:
/generate prompt:anime character, female warrior, fantasy armor, detailed
  1. Quick Refinement (Qwen):
  • Click ⚡ Qwen Edit
  • Prompt: "add magical sword in hand"
  • Steps: 8
  1. Polish (Flux):
  • Click ✏️ Flux Edit
  • Prompt: "enhance details, add glowing blue eyes"
  • Steps: 25
  1. Upscale:
  • Click 🔍 Upscale
  • Ratio: 4x
  • Denoise: 0.5
  • Steps: 20

Workflow 2: Landscape Art

  1. Generate Base:
/generate prompt:fantasy landscape, mountains, sunset, ethereal lighting width:1344 height:768
  1. Iterate Quickly:
  • Click ⚡ Qwen Edit
  • Try: "add castle on mountain peak"
  • Steps: 8
  • If not perfect, click again with variations
  1. Final Quality:
  • Click ✏️ Flux Edit
  • Prompt: "enhance details, add dramatic clouds"
  • Steps: 30
  1. Animate:
  • Click 🎬 Animate
  • Frames: 161
  • Prompt: "slow pan across the landscape"

Workflow 3: Quick Concept Iteration

  1. Start Simple:
/generate prompt:modern logo design, minimalist
  1. Rapid Iterations with Qwen:
  • Click ⚡ Qwen Edit: "make it blue instead of red" (8 steps)
  • Click ⚡ Qwen Edit: "add geometric shapes" (8 steps)
  • Click ⚡ Qwen Edit: "simplify the design" (8 steps)
  • Each edit: 30-60 seconds
  1. Final Version:
  • Once happy, use ✏️ Flux Edit for final quality
  • Then 🔍 Upscale to 4x

📊 Status and Information

Check Bot Status

/status

Shows:

  • Bot version
  • Discord connection status
  • ComfyUI connection status
  • System information

List Available LoRAs

/loras

Shows:

  • All available LoRAs
  • Organized by model type
  • How to use them

Get Help

/help

Shows:

  • Command reference
  • Parameter information
  • Quick tips
  • Links to documentation

💡 Pro Tips

Generation Tips

  1. Prompt Engineering:

    • Be specific and descriptive
    • Use style descriptors (photorealistic, anime, oil painting)
    • Add quality keywords (detailed, high quality, 8k)
    • Separate concepts with commas

  2. Parameter Optimization:

    • Start with defaults
    • Adjust one parameter at a time
    • Save seeds for good results
    • More steps isn't always better (20-30 is usually enough)

  3. LoRA Usage:

    • Start with strength 0.7-1.0
    • Lower strength (0.3-0.5) for subtle effects
    • Higher strength (1.2-2.0) for stronger influence
    • Check which model the LoRA is for

Editing Tips

  1. Choose Right Tool:

    • Testing ideas? → Qwen (fast)
    • Final result? → Flux (quality)
    • Multiple iterations? → Qwen first, Flux last

  2. Edit Prompts:

    • Be specific about the change
    • Mention what to keep: "keep everything else the same, just add..."
    • Describe desired result, not the process

  3. Steps vs Speed:

    • Qwen 4-8 steps: Maximum speed
    • Qwen 8-12 steps: Balanced
    • Flux 15-25 steps: Good quality
    • Flux 30-50 steps: Maximum quality

Upscaling Tips

  1. When to Upscale:

    • Only upscale finals (saves time)
    • Upscale after all edits
    • Good for prints and high-res needs

  2. Denoise Sweet Spots:

    • Photos: 0.3-0.5
    • Art: 0.5-0.7
    • Heavy enhancement: 0.7-1.0

  3. Ratio Selection:

    • 2x: Quick social media
    • 4x: Standard printing
    • 8x: Large format, posters

Video Tips

  1. Frame Selection:

    • 81: Social media loops
    • 121: Standard animations
    • 161: Presentations, dramatic effect

  2. Animation Prompts:

    • Simple movements work best
    • Mention camera movement
    • Avoid complex scene changes
    • Test with 81 frames first

  3. Performance:

    • Videos take 2-10 minutes
    • Don't spam video generation
    • Generate videos during off-peak

🚀 Quick Reference

Fast Iteration Loop

  1. Generate → Qwen Edit → Qwen Edit → Flux Edit → Upscale

Quality Priority

  1. Generate with high steps → Flux Edit → Upscale 4x → Done

Speed Priority

  1. Generate → Qwen Edit → Done (Total: 1-2 minutes)

Full Production

  1. Generate → Test with Qwen → Final Flux Edit → Upscale 8x → Animate

🎨 Ready to create? Try these examples in your Discord server!

See Features for detailed parameter information and User Guide for complete documentation.