Beyond DALL-E: Advanced Image Generation Workflows with ComfyUI
summary
ComfyUI workflows combine Stable Diffusion XL, LoRA and plugins to achieve precise results: demonstration pipelines run on runpod.io with Docker and a GitHub repo of workflows; examples include LoRA for style transfer to produce Ghibli-like images, a ControlNet + low-denoise image-to-image pipeline that turns the Python logo into a top-down island, Segment Anything + grounding to auto-mask and inpaint grapes out of a fruit still life, and a latent-space “flip” pipeline that successfully blends two prompts into an upside-down illusion.
AI-written summary · about
chapters (10)
- 0:00Introduction to ComfyUI
- 0:44Understanding ComfyUI
- 3:05Building a basic pipeline
- 8:52Styling with LoRA
- 11:31Advanced control techniques
- 16:02Combined generation workflows
- 17:11Inpainting and segmentation
- 20:24Low-level operations
- 23:28Conclusion and resources
- 24:19Q&A session
related talks
52mThe Yoga of Image Generation with Stable DIffusion and ComfyUI
01h 27mImage Generation with AI
47mAI Image Generation: Visual Artistry with Midjourney, Adobe Firefly, and DALL-E3
29mIntegrating Stable Diffusion Generative AI in Artist Workflows
25mDecoding Algorithmic Artistry: Visual Generative AI
28mIndustrial Strength DALLE-E:Scaling Complex Large Text & Image Models
34mHalf Hour of Flower: Planting the seeds for great Image Quality
49mAstronaut Horse: Stable Diffusion Artist Collaborations