Analog video synthesis, synthesized by machine learning.
This Lora was trained on 50 highly stylized images.
Workflow:
https://drive.google.com/file/d/1x2IHDaA-2GydtAAnGoVxZEPx8oeLHbOJ/view?usp=sharing
Settings:
Lora Strength: 2.5 is the sweet spot. 0.8 is the lowest that the Lora will show effects, 3.0 is the highest before it's effects generate diminishing returns.
Flux Guidance: 2.5
Scheduler and Sampler Combinations
-Euler & Simple: standard
-Huen & Normal: standard
-Huen & Beta: standard
-dpm_fast & ddim_uniform: more abstraction, color compression that resembles vintage TV texture
Max Shift: 0-1
Base Shift: 0-1
This Lora Pairs nicely with my 8mm Film Lora at a strength of 1.8 and other Loras trained on analog and glitch data.
Please share your results, feedback, and questions.
Happy Synthing!
Description
Version1
FAQ
Comments (9)
Hi the preview is looking amazing, i am dying to try this but having a hard time setting up Flux, many models I am missing out it seems. do you have download links handy? appreciate you!
Can you specify what models are missing?
actually i download every model needed from this page - https://civitai.com/models/617060/comfyui-workflow-for-flux-simple-lora but this workflow is showing just normal images, not synthesized images hmm wonder what I am doing wrong
this is what i am working with right now! lmk if you have any advice - https://drive.google.com/file/d/1N08ETb489j6UaqHUsogt1Zyo7B30pqlN/view?usp=sharing
@du1ce just checked this workflow, make sure you will have the Synthwave Lora downloaded and loaded into a lora loader. Here are my recommended settings: Lora Strength: 2.5 is the sweet spot. 0.8 is the lowest that the Lora will show effects, 3.0 is the highest before it's effects generate diminishing returns.
IN YOUR PROMPT YOU MUST INCLUDE "SYNTHWAVE AND VIDSYNTHSYNTH"
Flux Guidance: 2.5
Scheduler and Sampler Combinations
-Euler & Simple: standard
-Huen & Normal: standard
-Huen & Beta: standard
-dpm_fast & ddim_uniform: more abstraction, color compression that resembles vintage TV texture
Max Shift: 0-1
Base Shift: 0-1
Also when you write prompts, you have to give the model something to shape into the texture. If you just write "mushroom" it has to pull the analog texture out of thin air which it struggles with. Try something like A hyperdetailed mushroom character with humanoid features, its cap glowing bioluminescent patterns, blending neon hues with glitch-like distortions. The stem intertwines with refracted waves, set against a pulsating glitch synthwave background, merging organic and digital aesthetics. Synthwave, vidsynthsynth, video synth
@Calvin_Herbst Yasss it worked out :) it was the prompt problem that I didn't include the words. another question! does this work as Video too? if so - what would be needed?
@du1ce As of now, there is no video model for FLux. You could take the images into an image to video system like Kling of Luma. If you want to stay open source, you use stable diffusion1.5, but this model is flux only. I tried to train the model for SD but paused the work when I didn't like the results. In the meantime, I would recommend this for a similar glitch model in SD with animate diff but it's not as pretty as flux. https://civitai.com/models/189013/Retro%20Glitch%20-%20Style *make sure to download the SD15 version
@du1ce and make sure to share the images on the Lora's page, I give buzz to everyone who does.
@Calvin_Herbst whoaaa looks good! i will try this, thx!!!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.