Will do. Thank you!!

Great to hear! Thank You.

Glad it was helpful! And thanks for watching.

Thank you!!

Thank you! I am glad the video helped! If you have any questions during practice, feel free to ask. Happy to help, if I can.

Thanks for watching! Appreciate the support.

😊 Thank you!

Thank you!! There is no lcm model for stable video diffusion. Stability ai has only svd xt and svd models available. One does 25 frames and other 14 frames. Lcm Lora might work with animate diff, I doubt for svd atm.

Thank you!! I was experimenting with some other things, but that's a good idea. Will give it a shot. Just a tip: If you have 24gb vram you can up the resolution to full hd 1920x1080 and SVD can handle it. Currently there should be one or two upcoming workflows which include motion control and temporal ControlNet.Will see if inpainting works somehow. Also SVD 1.1 has been released. You have to download the model from hugging face. This model is specifically trains for motion bucket 127, 6fps, 25 frames, standard resolution. You should get consistent results from it.

Reduce the frame rate of the output or add frame interpolation in between. Techniques are shown here kzfaq.info/get/bejne/qMyYaLijlrvZoIk.htmlsi=jOtvZhmzxeLM5NIU

Hi, thank you! I will put animate diff / sparse Ctrl in the Todo list. Haven't played around with it yet.

@@controlaltaithat sounds great will be looking forward to it :) Thanks

Hi, video frames is the total frames for generation. FPS is how many frames per second it will generate. Assume say we have no restriction on video frames. In such a case say I need exactly 1 min of video. So 1 min of video can be in the following ways Sluggish: 1800 Frames but 30 fps = 60 seconds. Calculation goes as 1800 ÷ 30 = 60 seconds. Now I want it smooth, so I need to get it at 60 fps. That means 1800 Frames 60 fps = 30 seconds To get 1 min frames have to be 3600. That's 3600 ÷ 60 = 60 seconds. The reason to not play around this is because the model is trained for 25 frames only. So if you select 25 fps it would be a 1 second video. Hence default is kept at 6. You can then make a 2 to 5 second video playing around with fps value. But it will make the video more sluggish or smoother. Total frames are too less to increase or decrease fps. For the output after it has generated the frames to make it slightly smoother you say put 10 fps. It will take the generated video and run it slightly faster. So say you took 3600 frames generated at 60 fps which made a 1 min video. Now this is generated. The video combine node can make it run at the same speed of 60 fps, slower at 30 fps or faster at 120 fps without regeneration. Like playing a video of KZfaq at 1x speed, 0.5x speed and 2x speed.

@@controlaltai Thanks!

30 fps ? 9 second de video ?@@controlaltai

@@tvanime6747 I just gave an example. Svd only supports 2 to 5 second videos, max 25 frames only. Default 6 fps will give you approx 3 seconds.

Double click the left mouse.....that brings up search.

Try searching for "imageresize" and "image resize". One of them is the correct one.

Hi, thank you for the support. I Never used google collab so I cannot test it. Try this Please: ! add-apt-repository -y ppa:savoury1/ffmpeg4 ! apt-get -qq install -y ffmpeg

Hi, what is your motion id, fps and augmentation value in svd conditioning. Also what is your min cfg and the final cfg in ksmapler.

Just try with 16 is to 9, cause for me it worked for that. I could not get it to work for 9:16. Use low motion.

That can be simply done using the same seed and inpaint. You can use standard in painting, ip adapter or a detailer. All would work.

thx~!!@@controlaltai

Explain by what you mean can't load show text. Does it mean the nodes shows by its empty? If that's the case then its normal since you have to run queue prompt to get the info in the node.

​@@controlaltai At 36:38 is shows launching "ShowText", but this node is not in the list even though I installed and I loaded "Pythongosssss". I tried to making just the "ShowTest" my own custom node and removed the snake. That worked, and now I have "ShowText". But now when I run "Queue Prompt" i get a new error. "ERROR:root:Failed to validate prompt for output 6:... ERROR:root:* SVD_img2vid_Conditioning 3:...ERROR:root: - Required input is missing: init_image...ERROR:root:Output will be ignored

Check the workflow again and check the nodes connections. Something is wrong with the node connection. The error show SVD conditioning node is not receiving an image input.

Search comes from double left click. Try that and search for vhs.

@@controlaltai perfect, but VHS module is not present (but correctly installed, when I search it in the manager, it say to unistal or search for update)

Check if it shows up here. Right click - add nodes - Video Helper Suit, if it doesn't close everything and restart, still nothin then go to comfy manager - uninstall - restart - install again - restart. Restart means close browser and command prompt.

Cannot diagnose from just this error. Elaborte what are you doing and which node is giving the error.

A custom node, not the default one. Geth it via the comfy manager: palant - Image Resize for ComfyUI. You can then add node - image - "Image Resize" right at the bottom. Or Just search for "ImageResize" without space.

Yeah nothing is changed in SVD models, except you now get a even better mod SVD XT 1.1. Basically with SVD you don't have any control. If you want to do proper video animation you need to look at other tutorials related to "Animate Difff and IP Adapter". None of these are for people who never worked in Comfy. As you need to understand the basic know how in Comfy UI like, what is SD, ControlNet, Custom Nodes, How to go about in ComfyUI, shortcuts etc. First check out basic tutorials about comfy then come to these tutorials.

Not yet. Stability AI will release 3 official Motion LoRA for horizontal, vertical and zoom.

@@controlaltaithanks

@@controlaltai also anyway to do this as a batch?

Yes, in K-Sampler change "control after generate" to randomize. Then Below queue prompt tick mark "Extra". Change batch count to 3. This should generate 3 random videos effects form the same image and setting. You will get in consistent results.

No, 1.1 has a different setting sweet spot; the basic principle explained is the same, though.

Try these for 1.1 and consistent animations: Motion Bucket - 127 Min CFG 2 Target CFG 2.5 Augmentation (0 - 0.10) - mostly 0 Steps 40 Sampler Euler Ancestral Scheduler SGM_uniform

There are two. If you installed the custom node, you will find one as "Image Resize" and one as just "ImageResize". Try both to see which is the correct one.

@@controlaltai is this a default ComfyUI node or is it a custom node?

A custom node, not the default one. Geth it via the comfy manager: palant - Image Resize for ComfyUI. You can then add node - image - "Image Resize" right at the bottom. Or Just search for "ImageResize" without space.

The only way to extend or reduce time is playing around with the fps in the video output node. The model is designed for 25 frames at 6 fps only. Once you generate these 25 frames. How fast or slow you want can be via the fps in video combine node. You can add frame interpolation to smoothen it out .

Install this custom node form Comfy manager (Image resize by Palant), then when searching in add node, search for "ImageResize" without space. The one with space is the default one.

thanks @@controlaltai

Yeh I check. It's unfortunate. Another user reported this. The problem is with Torch, they dev says they plan to add support but even in torch 2.0 there was no support added. It will eventually come.

@@controlaltai is it working now?

@@SergeySokolov-rh7hu No way for me to test. I don't have a mac. I checked online, I could not find anything new regarding the same.

It’s already shared for Channel Members. Please check description on how to join.

@@controlaltai thanks, I build it from your video

Agree, I had animated Diff on the bucket list, was about to do a tutorial on that, and this came out. Will get to animated diff in some time.

A journey of a thousand miles starts with a single step, now a toy for kids under 6, and in a couple years a work routine for adults.

Quien lo use le saca el provecho, yo lo uso para documentales y mi cliente esta satisfecho

Its getting there. Give it some time. I totally get what you are saying. Since I am aware of what's upcoming (in pipeline), I can tell you things are progressing way beyond than expected. Problem is current hardware (consumer grade). We are like 1-2 years aways to make our own short 5 minute story driven animation, using nothing but single images.

@@controlaltai I hope it will be faster :)

Hi, 22:45 Explanation: I have 2 images (eyes open, eyes closed). I am trying to create an image sequence. For that I use a node called "Batch Image." The output from this node will combine the input images. So with just one batch image node I get 2 images (eyes open, eyes closed). I duplicated this node and daisy chained it in order to get 4 images in the sequence of eyes open, closed, open, closed. You then connect this output to SVD Conditioning. How does this Node Affect Output? SVD conditioning is predefined and does not have a text conditioning input. Instead of giving 1 image I try and give the SVD input multiple images in sequence to influence the animation output. Unfortunately, it is not possible to include the basics of comfy in every video as it is very vast. This tutorial does require some basic ComfyUI know how. Have a look at this tutorial to understand the basics: kzfaq.info/get/bejne/o7uhraRmqZbPaXk.html If you have any questions, feel free to ask. Happy to answer and help out.

@controlaltai That part I understood. WhereI got lost was the mouse jiujitsu you did right after you changed the image and got that one node that just says "Image" on it. Not really sure what you did to get that node, and what it actually does. I suspect it's like an "extension cord" like deal, but not sure. Sorry if I sound kind of thick here, but like I said, I am still very new to ConfyUI.

No problem. That's completely ok. I understand. Can you please tell me the exact time frames +-5 seconds. So I can check and give an explanation.

Okay I understood. That is called a re-route. Extension. You drag out from any node and in the middle of nowhere stop, leave the mouse it will give a reroute option. It's like an extension to connect longer noodles in the comfy graph. It's only function is extension for easy connection. The reason for fast forwarding was that's basic comfy thing. Important things are slowed down. But if you are new to comfy it's understandable. Feel free to ask any questions. The point of making these are you getting to use it and learn. Happy to help answer any further questions.

@@controlaltai Thanks for your explanation and you patience :)

It’s already mentioned in the video. Check here 3:07

@@controlaltai Anyway it seems it requre 4090, even 512x512 consumes whole 12 Gb VAM.