วีดีโอ

Thanks for helping me squeeze a few more frames out of my length

1st!

Thank you for doing this!

first

Plz make one video on how to use this Model on Google Colab or in another different cloud platform rather than locally bcoz eveeybody do not have high GPU, i uses Nvidia RTx Geforce 3050 4Gb, so it is so tought for me to run lagre models

You are so underrated. Thank you. I will stop using your mom as a lora. I promise.

First!!!1 :P

Time to turn the page on the Library saga.

Sending Snacks. I don't know the address, so I'll just send snacks to all addresses worldwide.

Cooking stuff as Usual, the comfyui community is on fire for last 15 days. i have installed my comfyui setup like 7 times in last 10 days, lots of development, lots of curiosity, lots of errors and bugs, lots of inexperience and inadequacy from my self. Thanks FiveBelowFiveUK for helping

nice but still way to much effort if you dont have any time left from the day anyway, something like noisee but in far better would be really awesome for some clips for musicvideos, doesnt even have to be edited together, just like giving the raw concept and then getting the needed scenes to edit them the old way, or did i miss something like this already existing? generating one by one and sorting out all the crap and headaches with the common video generators isnt it yet imo. beside the still ridiculous amount of prices and differences in looks

I'm going to dow land this once I get my new computer.

loving these new drama titles 🤣🤣

BatchMan 2: Return of the Framer.

I'm Batchman!

I was not able to join the stream live but I will watch it now gladly 👍

You are such an "expert". I'm not. Maybe you want to look into "it": Im confused because I just found out that some Flux Loras are working with Hunyuan Video. It depends on the lora's (i failed with any "style" lora yet, many did nothing or destroyed the video). The right prompting of Hunyuan to "integrate" them properly is key I guess. They can kind of degrade the video output quality a bit... But with V2V and another denoise/upscale step this doesn't matter much. I don't understand this. You'll have to explore please :D

The Legend of the White Cat.

1:10:35 "Now... pull... the... other... one". (as in pulling my leg, as in "you're joking")

Yeah, I asked an LLM "What is this word they say in this song? is it (insert word here) ?" "Yes (same word here) is the word". Then i looked up lyrics and it was 100% wrong and nobody else on the internet thought they said it was this word. It was just agreeing with me.

Can Parappa, wrap the wrapper with a rap. Can't let you do that Star Fox.

chatGPT suggested this video🔥🔥

3:10:50 easy enough to read.

Hello mister, could you share how to create a talking avatar like yours?

I find that individual prompting inpainting to upscale sections at a time, works best, but is insanely tedious and takes forever.

This stream got me to finally settle the question I had. Depreciated: Means to lower in value, often in relation to physical assets or financial situations. For example, you might describe a car as depreciated. Deprecated: Means to express disapproval or obsolescence, often in social contexts or tech-related discussions. For example, you might describe a software function as deprecated if it's discouraged from use but still works. I was always wondering if I was reading it wrong or it there really were 2 different words, and now I know there is. For python AI files, both words would basically mean the same generalized thing, "old and no longer the best to use".

Keep up the great work. I hope this video gets the views like the last video! WOOOOOOOOOOOOOOO LET'S GOOOO!

Good video.

I have something i can't find a solution: "...a node is missing the "class type property"..."

My best possible quality vs speed output (with a 4090) is reached with doing a 7 Seconds "low resolution" first, to get something that is perfecly in this "preview". To see it's not "garbage". With 432x240 resolution and 30 Steps. With Fast Lora I do it with only 15 Steps. So it could took 2 or 1 minute for this first "preview". Then the V2V for the final 848x464 resolution is done.... with only 50-70 denoise (could take around 4-6 minutes). I upscale this to 1.696x928 with the RealEsrgan2x (1 Minute), and then a frame interpolation with x3 and 60 FPS (2-3 minutes) video output. This is how I like it done. because it's still slow, but could be the right trade for quality vs speed :) Those Results could be extreme good.... But sometimes they suck, because of to less or much denoise in Vid2Vid. Then I just do it again with other denoise and/or another seed. I made some clips that looked like it was just real TV footage, and it's just slightly below FHD. I love this Arnold Schwarzenegger Lora, this gave me insane results, doing Terminator or Conan like scenes :D

star st-st-st- start y gin your start en your engine start your engine.

Looks great, but there is a second between 15:29 to 15:31 in the video that has glitching/distortion. Keep up the great work.

Okay question answered.

Starting my New year out with the RIGHT info! thanks 5B6UK!

You are good VJ.

Your videos are always awesome

Nice

Still no offline udio😢

1:30 🎶 🎵 💜

Yeah, very intresting stream

Can be any made for free .so not with paid website ....? ❤

Is that a chrome extension that is giving you colored tabs (comfyui, UE5, TH-cam, AI SAA…)?

noob question - notice that only 1 model is connected to the merge group sampler - is that how it supposed to be? looks like it allows to connect only one model, does it mean that flipping them around simply switches the 2 models from 0 to 1 respectively?

Seeing you trying to refine the captioning for lora training, makes me wonder if we could use flux redux output conditioning for lora training or not? I haven't done lora training since colab was no longer free to train lora, so I am not familiar with flux lora training. Also wondering could the combination of flux redux and condelta achieve what you are trying to get?

florence? what's the deal with this...why florence?

thanks for sharing

Hi! Is this tool only capable of working in the resolution of 576x1024? Is there a way to have any other resolution and aspect ratio? Maybe some workaround for that?..

Thank you so much 🙏❤ your workflow works well with my 8gb vram card

What a cool set of experiments!