Create shuffle-dance_workflow.md

shuffle-dance_workflow.md

@@ -0,0 +1,21 @@
+Source: https://www.youtube.com/shorts/SJKl6PEXklU
+
+Used instructions from Tokyo_Jab here. Used Temporal Kit for converting video to grid and back to video (with EBSynth). Below are the steps I took to convert this video.
+
+Removed background.
+
+Resized the video to 1024x1536. This is a multiple of 512x768.
+
+Used Temporal kit to generate the grid using values (Sides: 4, Height Resolution: 3072, frames per keyframe: 4, Check EbSynth Mode, Check Batch Settings > BatchRun)
+
+Used img2img with settings (Steps: 25, Sampler: DPM++ 2M, CFG scale: 7, Face restoration: CodeFormer, Size: 2048x3072, Model: edgeOfRealism_eorV20Fp16BakedVAE, Denoising strength: 0.8, Version: v1.2.0, ControlNet 0: "preprocessor: softedge_pidinet, model: control_v11p_sd15_softedge). This can be done with txt2img, I used img2img to run in batch.
+
+Temporal-Kit>EBSynth-process, Use read_last_settings, set output resolution to 1536.
+
+Run img2img on the keys directory to add more details, I used values (Steps: 25, Sampler: DPM++ 2M, CFG scale: 7, Face restoration: CodeFormer, Size: 1024x1536, Model: edgeOfRealism_eorV20Fp16BakedVAE, Denoising strength: 0.1 )
+
+Run EbSynth. I broke it into batches of 20 keys. Temporal-kit has the option for doing this when generating the grid. I found that it results in incomplete grids and that breaks the consistency between images.
+
+Temporal kit > recombine ebsynth.
+
+Open video in DaVinci to use deflicker.