Completed
Selecting baseline freeU s values
Class Central Classrooms beta
YouTube videos curated by Class Central.
Classroom Contents
Improving Stable Diffusion Images with FreeU - Optimizing SDXL, LCM, and Turbo Models
Automatically move to the next video in the Classroom when playback concludes
- 1 Introduction
- 2 FreeU paper
- 3 How do b and s value interact
- 4 ComfyRoll nodes
- 5 CivitAI workflows downloads
- 6 ComfyRoll workflows
- 7 ComfyRoll Grid error
- 8 Debugging the grid error
- 9 Trying out ComfyRoll XY list
- 10 Finding a fix for grid error
- 11 ComfyUI API
- 12 ComfyUI API JSON
- 13 How to request ComfyUI API
- 14 Re-evaluating the API JSON file
- 15 Code overview
- 16 Selecting baseline freeU s values
- 17 Handling asynchronous calls as ComfyUI API doesn't have callbacks
- 18 Setting up code environment
- 19 Executing the code
- 20 Analyzing the dynamic b values for normal FreeU
- 21 Narrowing the search window
- 22 Why choose static s values of 1.1 and 0.2? Why not use 0, 0 or 1, 1?
- 23 Dynamic s values
- 24 Getting original seed value
- 25 Pointing out my mistake with my seed value wasn't working
- 26 Comparing with original generations as a sanity check
- 27 Comparing results so far
- 28 Workflow for kohya deep shrink
- 29 LCM workflow
- 30 Dynamic b values for LCM
- 31 Dynamic s values for LCM
- 32 Updated comparison of all results so far
- 33 Extracting SDXL turbo LoRA
- 34 SDXL Turbo LoRA workflow
- 35 Dynamic b values for SDXL Turbo
- 36 Comparing different b1, b2 values for SDXL Turbo
- 37 Dynamic s values for SDXL Turbo
- 38 Final comparison of all results so far
