ComfyAnonymous

Follow ComfyUI’s workshop from local image generation through guidance, latent compression, regional control and the engineering needed to turn workflows into applications.

• What should a generative interface let you control?
  0:14 ↗
• From a personal experiment to Comfy Org
  4:45 ↗
• Read the galaxy-bottle graph
  7:08 ↗
• Reuse the pipeline for video and sampling patches
  10:10 ↗
• What the two text-encoding branches actually do
  12:56 ↗
• Why the sampler works in latent space
  16:56 ↗
• Automating a workflow does not solve evaluation
  18:53 ↗
• Remote models, 3D and the boundary of core support
  23:07 ↗
• Virtual try-on and character consistency
  27:13 ↗
• When a reference image is not enough
  32:59 ↗
• Night at the top, daytime at the bottom
  38:12 ↗
• Local control includes responsibility for the output
  41:39 ↗
• Control, extensibility and the cost of learning
  44:43 ↗
• Make local workflows reliable before building the cloud
  47:45 ↗