First published at 19:38 UTC on February 1st, 2024.
"I got bored one day and i put everything on a bagel".
- IPAdapters chained and masked composited
- Animdiff v3 and gen2 nodes
- Face Swap and restoration
- 5 control nets that can all be mixed/matched bypassed
- An upscaler via Ergan through…
MORE
"I got bored one day and i put everything on a bagel".
- IPAdapters chained and masked composited
- Animdiff v3 and gen2 nodes
- Face Swap and restoration
- 5 control nets that can all be mixed/matched bypassed
- An upscaler via Ergan through pixel space,
- Hand MeshGraphormer
- Prompt Travelling
- Interpolation
This video is NOT a tutorial, but instead, an explanation as to WHY we're seeing a 'convergence' in methodologies as part of working with comfyUI and animatediff. Even Netflix picked up on the trend as they are now recruiting VFX people familiar with free and opensource tools such as AnimateDiff/ControlNets/Reactor and of course, IPAdapter.
00:00: Intro and what it contains
01:45: Don't pay for workflows!
03:45: AnimateDiff V3 Gen2 Nodes
06:20: Single frame generation
09:00: Thinking like an animator
12:20: Why I'm not using pipes here and Trung's 0246
15:20: This is not a 'filter'
19:20: Pre-Processor best practices
25:00: Depth map: Midas vs Zoe vs Zoe AD
26:00: Why I don't use a second Ksampler
27:00: Control nets and COCO masking Lineart
31:10: Keeping the background steady with tile controlnet
35:00: Where the Bagel is headed: automated background infill with a vision model (moondream)
38:00: IPAdapters masking with a composite mask
46:20: Segmenting anything to anything
47:00: Chaining IPAdapters
48:22: FaceSwapping with Reactor with restauration via GFPGAN
49:00: Warpspeed and Hand Meshgraphormer woes
53:00: Prompt Travelling and Parsec
56:00: More tools!
57:00: Upscale via pixel space
59:54: Interpolation and saving
1:00:00: Concluding nodes and Banodoco
WHY do we need background consistency? HOW do we obtain it? This is what i want to explore in this video, alongside concepts such as bypassing the issue of 'squished' clipvision images (which much be square) when dealing with vertical or portrait videos.
I'm releasing this video alongside my tutorial workflow which you can obtain for free (evidently, you should NEVER pay for workflows) on th..
LESS