วีดีโอ

Impressive content, CG Pixel. Can't wait to see your next upload from you. I hit the thumbs up icon on your segment. Keep up the fantastic work. Your detailed explanation of the wave speed node's impact on image generation was enlightening. Have you considered exploring potential optimizations for video generation with high-complexity prompts in future videos?

very good explaining thanks a lot! have you already tried this workflow with wavespeed?

Awesome, thanks a lot!

i don't have musgrave texture

how do you connect t2v to upscale?

My question is can you load LoRAs into specific areas and stop "bleeding". That would help a lot.

Very good teaching video, but why is the image I generated black? What should I do? Thank you. * Tried to change vae, but still generated black images

how does the vae guff file compare with the 327 mb ae.safetensors file?

Thanks, this flux inpaint work good even with Flux Schnell in 4 steps

How to add more time in the generation of the video (more than 3secs) ?

I am getting following error: Missing Node Types: Florence2Run Florence2ModelLoader ShowText|pysssss Int Literal UnetLoaderGGUF GetNode SetNode and many more... please help. ps:I have already added ComfyUI-Florence2 to custom nodes.

theres now a flux 1 dev hyper nf4 did you try that one let me know thanks for the information and video.

my outputs coming out great , just remeber to prompt this " best video quality of

Your learning :) nice video.

You forgot to include the Florence2 download in your video/description or as a required node. It also would have helped if you expanded it in the vid, but I was able to figure out that it requires "Florence-2-base" model via your workflow JSON Overall very helpful video though. Thanks!

What about coloring the lineart? Where can I download the flux forge model?

Hello, I'm new to using the tool, is there a chat so I can contact you to help me use the workflow?

Hello, I am from the SeaArtAI ComfyUI team. We would like to invite you to collaborate with us on the ComfyUI project. Could you kindly provide an email address so we can send you further details about the collaboration?

okay well you clearly know your way around ComfyUI... subscribed

Khouya enta les Zommes! Excellent delivery, consistently on point.

is ltx okay for making stylized or illustration style video like animated diff? i like ltx because it is fast just wondering does anyone have tried it for non-realistic style?

Even tho I installed the Florence 2 node, I still got the Missing Node Type message error that say Florence2ModelLoader were not found. Do you have any clue how to fix that?

Even tho I installed the Florence 2 node, I still got the Missing Node Type message error that say Florence2ModelLoader were not found. Do you have any clue how to fix that?

" FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers" got this error

Thanks for sharing. I havent been able to get 9.1 to work on my 6gb laptop. But ill keep trying to figure it out. Maybe your workflow might help.

I got this error "Unexpected architecture type in GGUF file, expected one of flux, sd1, sdxl, t5encoder but got 'hyvid'" ?

great works. !!!

Can you use a Lora of yourself added into ComfyAI currently being used with image creation for video?

*This is your pilot speaking

multiAreaConditioning node - is it still working? I cannot get it to work :(

@cgpixel6745 i tried same but not working failed at ksampler steps error all tensors need to be on same gpu

Great Video. Your pixelation problem might be related to mismatch in FPS. In T2V section you specified 24 FPS but in Upscale 30 FPS. I changed to 24 in Upscale section and everything looks good

Thanks, I hadn't realised you can do video to video with gguf. But of course you can feed the video into the latent space. However, I don't see where you are setting the denoise value. This would seem to be important, setting it too high will just ignore the input video, and too low will keep it too much the same. Anyhow, the simple workflow still works so is not an issue for me. I think the regime of trying to do everything all at once in a single workflow is a bad idea, but that's what most other youtubers are doing too. Better to keep a single workflow that does one thing, most of one's results are thrown away, and the ones you want to keep can then be thrown into rife or upscale, and that keeps the main process minimal.

Would it be possible to designate specific loras for each area? Eg. couples pictures where I can choose to use personal loras for each person. Thanks for the video, Subscribed.

Hey thats a great and quick video. I am not able to find find teh CLIPTextEncodeFlux node in the manager/ custom nodes, do you know what might be the issue?

Any suggestions after re-installing? This keeps giving me a conflicted nodes error for -- Florence2Model Conflicted Nodes: DownloadAndLoadFlorence2Model [comfyui-tensorops] Florence2Run [comfyui-tensorops]

can one set starting image for the text to video to be image to video instead? If one does not want to use generated text, but image itself as base

It's not true img2vid. It only uses the LLM to describe the image. It does not take the actual image and animate it.

I was getting the error "SamplerCustomAdvanced: Required input is missing: latent_image". I had to manually connect EmptyHunyuanLatentVideo.Latent to SamplerCustomAdvanced.latent_image. Is this the right way or am I missing something?

Your videos are always exciting.. Expect, They never work, lol

how to add loras to gguf

Great tutorial it was very easy to follow along! Thanks!

thanks man! quick questionhere, how to show gpu,cpu, ram etc workload percentage there? i have installed crystool but after the update few months ago it doesnt show on my comfyui

removal background is ok but not replacing background ; is that a problem of Load CLIP VISION? i have choice clip-vision-vit-h.safetensors

it doesn't work for me, which file is below "IP ADAPTER.SAFETENSORS" , i can't find it on huggingface or github..when i change background it stays black.

Workflow seems to get stuck on "Requested to load MochiTEModel_"(terminal) , and in GUI it shows it's on Clip Text Encode (Negative Prompt)

We need the new workflow version 🙏 TripleCLIPLoader does not work with the new comfy(v0.3.8) and flux, it generates black image. If change to DualCLIPLoader, it will work but lose a dimention. How to solve this?

hello, thanks for yout job.I have 2 problem: 1/ i get this message failed "KSampler index is out of bounds for dimension with size 0 " and the node UPSCALE IMAGE BY has the cricle red on the paramzeter IMAGE ?? what can i do ? Isse i have the same prompt that yours but you have no faile with this NODE UPSCALE IMAGE BY 2/ What sort of mp4 i have to find in the node LOAD VIDEO (PATH) ?

Whats your gpu? what vRam?

Great video! I like that you added prompt generator and frame interpolation. That makes a huge difference. Ai upscale would be great too! Thumbs up!