Need help? Check out our Discord channel: discord.gg/ztHKU2bgsD I've added some solutions and tips, the community is also very helpful, so don't be shy to ask for help 😉

@bhabasankardagar5810

3 ай бұрын

Thanks, It works with me select_every_nth : 15 and 480-1080, but it is taking too long, I have CP Config with 20GB RAM, Core i3, and Win 11. Let me know if there is any process to fast, I want to create 20-second video, can I upload the image segment in "Keyframe IPAdapter -Load Image" to speedup process?

@bhabasankardagar5810

3 ай бұрын

It's taking too long to create videos, so I'm considering generating animated sequence images instead. I'll merge these sequence images using Premiere Pro and create the video myself.

@MDMZ

3 ай бұрын

@@bhabasankardagar5810 the process relies heavily on your GPU's VRAM

@MDMZ

3 ай бұрын

@@bhabasankardagar5810 that's a good workaround

@devonandersson300

2 ай бұрын

@MDMZ Can confirm as of 12/03 that following your tutorial steps work perfectly. Was not a ComfyUI user (InvokeAI) - but I needed a solution that can work with video. I will try to combine it with the new 4 step SDXL Lightning or JuggernautXL Lightning models. Seem a PERFECT fit for good quality vs speed IF it works.

Your tutorials are one of the best and even beginners can become almost like pros by seeing your videos 🙌🏻

@MDMZ

3 ай бұрын

Happy to hear that!

Wow this is a great tutorial. It's taking its sweet time on my PC LOL but none the less it is actually working! I've seen so many vid to vid confyui videos, and everyon is jumping from left to right, with no coherency, no explaination about what model, and nodes do what, thanks for being super clear about those things. You single handedly just made this whole thing easy!

@MDMZ

3 ай бұрын

that's really great to hear. Thank you 🙏

@Aryannnnnn217

Ай бұрын

Will it work on rtx 3060?

@randy2d

Ай бұрын

@@Aryannnnnn217 well mine is a 3060 but the 12 gig version, also I kinda improved a few bits on the workflow and it is actually really good now

@Aryannnnnn217

Ай бұрын

@@randy2d mine is also 12 gig version, but i just shifted to comfyui, in A1111 my 3060 couldn't do controlnet and hires fix in sdxl models.. so im wondering if this workflow will work on my system? thanks for reply

@randy2d

Ай бұрын

@@Aryannnnnn217 I never ran videos higher resolution than 960x512 because the upscaler I use I can just set the size I want to upsale to and than send it to the video combine to export

Thanks for the video! Most creators forget, to show which models they got and where to put them in the ComfyUi folder. This step by step video helped a lot.

@MDMZ

3 ай бұрын

glad it was helpful

Excellent explanation! Kudos bro... You deserve millions of subscribers!

@MDMZ

3 ай бұрын

Thank you so much 😀

Incredible man!! Thanks, I was waiting for this! I prefer Comfyui than deforum. you are the best!💪

@MDMZ

3 ай бұрын

Glad you like it!

4:20 after download everything. Thank you for amazing tutorial!!

@MDMZ

2 ай бұрын

Glad it helped!

One idea to improve vídeo background. Try remove background first. Then apply a specific node only for background generation to avoid flickering. if you see flickering on hands you may use a technique by creating a boundering box that stylizes only hands and uses any hands detailers tools (lora, or node).

@MDMZ

3 ай бұрын

great tips!

the future of art... downloading the newest hard to find files Thanks for the tut, it was helpful I don't know how I would have figured out all those steps

@handlenumber707

Ай бұрын

Think of it as modern day gold mining, nothing more than just another scheme to get people to hand over their ideas for free.

You are amazing ya akhi 😎 as always awesome and creative videos 👍🏻❤️

@chouaibphenix1082

4 ай бұрын

frr mahabtch tslahli ma3lich fb tfhmni kimah psk dert kifo mhbtch tmchi mlwl

@MDMZ

3 ай бұрын

🙏

Best lesson ever. It's a pleasure to listen you)

@MDMZ

3 ай бұрын

glad you liked the video

@handlenumber707

Ай бұрын

It's interesting, but without even paying much attention I can tell it takes a level of involvement comparable with traditional methods. Until so called generative AI offers simplistic prompting, nothing changes. You'll end up having to pay experts to use these systems. I see no benefit to anyone apart perhaps for those owning severs, sifting through endless input codes, searching for some kind of pay-dirt. It's a hard ask. A.I. systems (a fad) will NOT replace traditional techniques.

@MDMZ

Ай бұрын

@@handlenumber707 You didn't just write "you'll end up having to pay experts to use it" and "i see no benefit to anyone" in the same sentence 😅

@handlenumber707

Ай бұрын

@@MDMZ Wasn't the whole idea to avoid paying people to do things?

Thanks for the clarity.

@MDMZ

3 ай бұрын

Glad it was helpful!

Such a useful video, thanks heaps for putting this together.

@MDMZ

3 ай бұрын

Glad it was helpful!

your the best sir

very nice and love using AI for animated features. Great sharing

@MDMZ

3 ай бұрын

Thank you! Cheers!

thnaks mdmz for yr effort :)

Thanks so much for the tutorial! just wondering how to keep the character and background to be consistent and a bit stable? I had kept similar setting multiple times but the outcome of human and background still changing a lot.

@MDMZ

2 ай бұрын

you can try playing with the main settings I mentioned. there's no exact formula, so try different combinations, I've tried to explain what every one of those settings does in the video

Good information thanks 😊

Hi! Thanks for the tutorial. One question, which controlnet model are you using? depth, openpose etc...thanks.

@MDMZ

3 ай бұрын

hi, it's depth, as mentioned in the video 😉

Thank you very-very much !❤️👍👍

@MDMZ

2 ай бұрын

You're welcome 😊

Buckle up, creators! This tutorial featuring ComfyUI IPAdapter + HotshotXL is your ticket to a whole new dimension of video wizardry. Transform your content with the power of A.I., and let the magic unfold! 🌟🤖

@handlenumber707

Ай бұрын

Let me get this right. You take a video of someone moving around. Then you upload the video, paying money to use this service. Then you type in some prompts, and you get an animated character back? You can do this for free on your own computer, without the middleman, and without sharing your ideas. It's called motion capture.

brilliant

seeing this awesome stuff will eventually wear down my distaste for node based systems, still clinging to A1111 for now, as an aside, this same reason I cannot abandon Carrara 3D for Blender 😛it even uses it's lovely shader tree in Octane render keeping those awful nodes hidden unless I choose to torture myself with bricks and spaghetti

@MDMZ

3 ай бұрын

I was scared of it at first too, just like any other tools, ComfyUI gets easier the more you used it

Thanks for sharing looking forward to trying it out!

Thank you. 감사합니다.

Great tutorial! Thank You! Could You please help me with one thing? My "Video Combine VHS" nodes are missing video formats - only "image/gif" and "image/webp" is available. WHat did I miss?

@MDMZ

3 ай бұрын

are you using the same workflow from this video? in any case, you can export to webp and convert later

@user-kx5hd6fx3t

3 ай бұрын

the node "animatediff combine" change to "video combine" . there are two nodes look like the same one . but different. try it again.

Hey really nice video i have watched it like 10 times on the last month. I have a question, is there a way to only animate the character but keeping the background static? Would be really awesome

@MDMZ

3 ай бұрын

you can rotoscope the subject out of your video first, and run it separately from the background, there may be a way to do it directly on ComfyUI but I haven't looked into it

great tutorial! well done! In the Video Combine window, I don't have any video formats like your video, only 2 image formats, do you know what that's about?

@MDMZ

2 ай бұрын

make sure you update ComfyUI and all the nodes

Thank you for the video. Is there a SD1.5 alternative of HotshotXL?

@MDMZ

2 ай бұрын

yes, mm-Stabilized_high is a good one

Thank you very much 👏🏼 One question. Is it possible to make a 15-minute video like this? Or is it only suitable for short videos of a few seconds? Thank you in advance

@MDMZ

2 ай бұрын

I haven't tried a video that long, I haven't encountered restrictions on video duration, but a 15 mins video will surely take so long to process if it works, why not give it a shot ?

@shizool2359

Ай бұрын

15 minute or 15 sec .15 minute destroy ur pc bro😂😂😂

thank for ur great tutorial.is there any limition for frame rendering? i use ur workflow for a 32 seconds video file and its like 30 frames(1000 png) and i got this error after 1 hours render time on my 3090ti: numpy.core._exceptions._ArrayMemoryError: Unable to allocate 6.43 GiB for an array with shape (976, 1024, 576, 3) and data type float32

@MDMZ

3 ай бұрын

probably running out of VRAM, make sure your GPU is not being overused by other apps during the process

Great tutorial. Can we do sizes like 1920x1080 and how long would that take ie 5-10 seconds. Is there anyway to have it create a sequence instead of a mp4 incase it fails to continue?

@MDMZ

2 ай бұрын

you can definitely go with other resolutions, time is almost impossible to predict, give it a shot

great!! thank you! to integrate a selfmade lora file: is it the best way to put it between the checkpoint and positiv text prompt or what would be your suggestion? thx in advance!

@MDMZ

3 ай бұрын

hi, tbh I'm not entirely sure, I will need to look into it

Thank you for your clean and helpful video. I tried to run this on my local machine but unfortunately I do not have enough vram. Do you have any recommendation on any cloud service?

@MDMZ

Ай бұрын

Thinkdiffusion is one of them, but I can't guarantee that all nodes are available on online services

hi, thanks for the tutorial it was a great help for a beginner like me. how can I Add my own custom SDXL Lora to the prompts here? like where do I connect em? thanks in advance

@MDMZ

2 ай бұрын

This would be a seperate tutorial on its own, did you try finding other videos on youtube ?

Thanks a lot

thanks for helping

Tx 4 video. One question: is it possible to have different animations but with the same character that I'm designing? If I filmed my little reference videos to animate my character, I can have this character with these sequences filmed for a short film ¿?

@MDMZ

Ай бұрын

the best way to get the same character is to train a model on a set of images of that character

@seosevilla

Ай бұрын

@@MDMZ Ok. Have you or know good one video tutorial for that¿? Many thanks

Thanks for the video! But why do I have a lot of paper cranes on my background, the original video is clean white, how do you make sure that the background is rarely affected

@MDMZ

2 ай бұрын

that shouldnt happen, try reinstalling comfyUI, could be a software issue

Is it alternative of warpfusion because I was going to buy that one, should I use this or wrap fusion

@MDMZ

3 ай бұрын

I find this much more consistent, warpfusion is getting better too

Do you need a video card for this? or can it run on Google Colab? Thank you

@MDMZ

3 ай бұрын

for this method, you need a video card, if you don't have a decent one, you can run it on the cloud(for a fee): kzread.info/dash/bejne/ioSGusqnotDgqtY.html

First of all thank you for your efforts for this great information and video. i am a mac user. i am using m2. zsh: killed, TypeError: Failed to fetch. does this mean that the RAM is not enough? What are the minimum computer specs I should have. I would really appreciate your help. Sincerely,

@MDMZ

3 ай бұрын

Hi, I can't tell for sure, but I know that it's challenging to make this work on Mac, have you followed the installation guide for MAC on the official guide ? and also, what's the full error text

Man you are awesome, thanks for your time and effort❤❤❤ do you know is it possible to use multiple controlnets in this pipeline? Depth+edge detection? I tried to use multi controlnet node but I got error with ip adapter then😢

@MDMZ

3 ай бұрын

theoretically, it should be possible, I haven't tried it myself.

@bboyhafan

Ай бұрын

i tried to with multiple control net, but it work just with about 20 frames, but when i try to make more frames of video there is error

excelente🥥

bien explicado felicitaciones solo una duda , me sale: When loading the graph, the following node types were not found: IPAdapterApply Nodes that have failed to load will show as red on the graph.

@MDMZ

2 ай бұрын

check the pinned comment, just shared a fix

@AlessandroMDC

10 күн бұрын

@@MDMZ gracias ahora si me salió

Thanks for sharing. I have a question I would like to ask. What are the minimum requirements for a graphics card?

@MDMZ

Ай бұрын

I recommend atleast 12GB of VRAM

I am on Mac and this seems to be PC windows only .... but interesting to know about its existence. How would you rate this tool for video stylisation/transformation compared to RunwayML video to video?

@MDMZ

Ай бұрын

this in my opinion is much better than RML, there are ways to run it on MAC, but it's huge difference when using an NVIDIA GPU

Thank you very much; you explained it really nicely. I'm currently at 50% and curious to see what comes out of it :) PS: Can the same thing be done with plants? For example, modeling how a plant grows and continues to change over time?

@MDMZ

2 ай бұрын

have fun! I'm really not sure about your question, sounds like you're talking about generating video from scratch?

Hi, please I need your help, I just updated the comfyui, did update all, and I lost apply ipadapter within the video reference, and also the ipadapter from the keyframe adapter section.

@jamilmalas

2 ай бұрын

I just found your comment about the update, thanks a lot, شكرا يا حبيبي

@MDMZ

Ай бұрын

u r welcome, glad it worked

great video, I just want to know how could I train my own model data from my own art set 2:30 and use that as the style reference?

@MDMZ

2 ай бұрын

yeah you can do that, I don't have a video on training your own model, but there are several tutorials on youtube

Can you share how much time did it take to train and GPU used? I used A6000 for several hrs suddenly saw Kill.. (May be my computer went to sleep) I should have chose 16 versions, Is there anyway to save progress on each steps? Thank you so much!

@MDMZ

Ай бұрын

what do u mean by 16 versions ? Anyways, it's normal for this process to be a little slow, if it stops running, check the cmd window for errors

In order to Apply IPAdapter, can you provide the reference image of Origami

This guide is so cool. What do I need to change so it will have better result for celebs?

@MDMZ

3 ай бұрын

a model trained on celebrity pics would probably help, but sometimes using the person's name in the prompt works fine

@MrReefxl

3 ай бұрын

I have tried that, but I get many artifacts on face and cloths@@MDMZ

How long does it take to process that e.g. dancing video? Which card did you do it on?

Maybe a dumb question, but is this at all possible on Mac machines? I have an M3 Pro with 36G of shared RAM and would love to try this out

@MDMZ

2 ай бұрын

the installation process is different, technically, it works, but the power of an NVIDIA GPU is unmatched when it comes to AI processing

Is there any way to do it with image to video pose instead of prompt to video pose?

Thanks for the video! Tell me, is my 4070 video card with 12 gigabytes of memory suitable for this configuration? Because according to your configuration, the video memory is fully loaded and the processing of the video for 100 frames stops at 5% of the progress.

@MDMZ

3 ай бұрын

12Gb should be alright, try rendering at a lower resolution, or lower steps

I have only one problem with this workflow: Following all your settings, the time from when I hit "queue promt" until I see my first frame is too long. What settings can I tweak to make it faster, without affecting it too much? When I find the look I like I'll go back to the high settings.

@MDMZ

3 ай бұрын

you can try setting the extract_every_nth to 10 or something higher, this way you'll process less frames and get to see what it looks like in much shorter time

Amazing Video. I love it. 😍 Bro, can you please make a video where I can change the character in the video to my character and transform it ? I am actually looking for somethings like this for long days.

@MDMZ

3 ай бұрын

great idea, I will look into it

@Cruse1914CODM

3 ай бұрын

Thanks a lot! You are so kind 🤗. I am very happy that you read my comment and replied. I will stay tuned.@@MDMZ

Nice 🤩😍

@MDMZ

3 ай бұрын

Thank u bro

Hey @MDMZ I have installed Comfy UI on my Mac M2. using another tutorial. Then when I came to go through your tutorial It seems I don't have access to the manager tab nor the share tab available to me on that window. I don't suppose you know why that is? Anyway thanks for sharing either way.

@MDMZ

2 ай бұрын

The whole tab is not showing or just the manager button ? Try updating comfyui, there's another solution for the manager button disappearing, in the pinned comment

@user-kd8oe1pk1m

Ай бұрын

@@MDMZ Just the manager button.

@user-kd8oe1pk1m

Ай бұрын

Ive have figured it out!

Does it need an external GPU to run ComfyUI IPAdapter + HotshotXL? Or it can be used without an external GPU?

@handlenumber707

Ай бұрын

It sounds painful.

Thanks for the tutorial bro! Yet i can't run Comfy UI from the batch folder. It shows an error about Nvidia old version and Cuda drive not compaticle with pytorch or something like that. Can you give me tip to solve this.

@MDMZ

3 ай бұрын

make sure your NVidia GPU is up to date, you might find more help here: discuss.pytorch.org/t/cuda-versioning-and-pytorch-compatibility/189777/9

Well it runs without any errors for me, but starting at the Video Combine Node all it generates is noise. Fractions of the prompt are even visible, but only so barely. Triple checked all models and settings but still it's like that. img2img same model is fine. ControlNet Preview Image seems fine. Any ideas?

@MDMZ

2 ай бұрын

this happened to me before, I solved it by reinstalling from scratch

great tutorial,...!!!! i wanna try, where i can find the original video of dancing man?

@MDMZ

15 күн бұрын

just added in the description

@azalaka22

13 күн бұрын

@@MDMZ thank you and it work ,.... thank you,...

nice one . .,how long it take u to render,. why in my setting in low v ram its too slow. although i have good gpu. ,2080 super

@MDMZ

Ай бұрын

if am not mistaken, the 2080 super has 8GB of VRAM ? which is considered a little low for this, you need atleast 12, it won't be blazing fast even with 24GB

@Dharmatography

11 күн бұрын

@@MDMZ thanks bro. Can you make that popular rose or plants dancing animation tutorial

I got this error: "could not be loaded with cv" pointing to the image_encoder. After downloading the encoder recommended in the IPAdapter Plus page I got it working. The link in the description points to a G model, while the IPAdapter Plus is an H model. Not sure if this is important, but it seemed to be in my case.

@MDMZ

4 ай бұрын

thanks for sharing, I will look into it, I'm using the same exact files in the tutorial and it works fine for me

Thank you so much for your amazing content! Your tutorials have been incredibly helpful. I followed your instructions in the video, but I keep encountering this error: Value not in list: ipadapter_file: 'ip-adapter-plus_sdxl_vit-h.safetensors' not in [ ] I’ve double-checked that the file exists and is in the correct directory, but I still can’t seem to resolve the issue. Could you please provide some guidance on how to fix this?

@nicholasescalante9004

20 күн бұрын

Hi ran into the same issue. IPadapter changed how it works. You need to move this safetensor and place it into "comfyui/model/ipadapter" instead. Create the folder if its not there.

@MDMZ

20 күн бұрын

you beat me to the answer, thanks for helping out!

@sallar7

19 күн бұрын

@@nicholasescalante9004 Thanks for the reply. Yes, I already saw the same explanation on Discord, and it worked. However, now I'm getting another error with "KSamplerAdvanced". Maybe I can send you a screenshot of the error, and you can help me fix it? the error says: Expected query, key, and value to have the same dtype, but got query.dtype: struct c10::Half key.dtype: float and value.dtype: float instead

Great tutorial man, but please next time tell us we need to install git from the official page in the first place xD

@MDMZ

3 ай бұрын

oh, I didn't realized it was necessary to do manually, may I know at which step you realized that and how you found out that u need to install it? I will pin the solution fore everyone else who runs into the same issue, thanks a lot!

@menwiththemask

3 ай бұрын

@@MDMZ Ehi there, I needed to install Git when I first run cmd and pasted the link

@musicspartan9905

3 ай бұрын

I thought I messed up on the first step, thanks bro

When loading the graph, the following node types were not found: IPAdapterApply Nodes that have failed to load will show as red on the graph.

@MDMZ

2 ай бұрын

check the pinned comment, just shared a fix

Hi my friend. I was trying to keep a similar face to video origin but I couldn't do it yet. Maybe trying another ip adapter?

@MDMZ

3 ай бұрын

looks up IP-Adapter-FaceID

i get this error when it gets to the sampler : Error occurred when executing KSamplerAdvanced: Expected query, key, and value to have the same dtype, but got query.dtype: struct c10::Half key.dtype: float and value.dtype: float instead. any idea what should i do ?

@MDMZ

2 ай бұрын

hi, you can check the pinned comment

I have a rtx 16go, 80go ram, and a i9 9900k, confy ui Works grate when i Just do pictures and its very fast, but with this workflow at the ksampler step it start to be very slow and take an hour to render in the ksampler. Comfyui start to be very laggy and freeze a lot at this moment… do you know why ??

@MDMZ

3 ай бұрын

that's unfortunate, seems like it's heavier on GPU VRAM, although 16 should be enough, you can try reducing the resolution, make sure your GPU driver is up to date, and make sure the GPU is not being overtaken by another application

I remember you mentioned a cloud site I could use to run Comfy since my PC takes forever. What was that site? I couldn't find it.

@MDMZ

2 ай бұрын

ThinkDiffusion

Hello MDMZ and thank you for this incredible video (as ever) !! Can any of you tell me why I don't have the "Manager" and "Share" options in my ComfyUI interface ? Thank you all !

@MDMZ

7 күн бұрын

hey, can you try to install the manager t using the steps in this video? kzread.info/dash/bejne/d5N4mdtpic2raJs.html

@lanknwa

7 күн бұрын

@@MDMZ I'll try it ! Thank you mate. 🫶

I tested your workflow and noticed, that it works only with Protovision checkpoint. Can you explain what unique specifies it has ? And what other chekpoints works with that workflow?

@MDMZ

2 ай бұрын

make sure you use an SDXL checkpoint, I tested it with atleast 5 models other than protovision, shouldnt be an issue

Sir help me, how to fix this Error(s) in loading state_dict for ImageProjModel: size mismatch for proj.weight: copying a param with shape torch.Size([8192, 1280]) from checkpoint, the shape in current model is torch.Size([8192, 1024])

@NEW-BLACKPINK

2 ай бұрын

Reply me

@MDMZ

2 ай бұрын

Hi, please check the pinned comment

@rhyswynn

2 ай бұрын

I had a similar issue, it was because I didn't select the image_encoder in the Load CLIP Vision node

Hi How can i fix that "When loading the graph, the following node types were not found: IPAdapterApply Nodes that have failed to load will show as red on the graph."?

@MDMZ

2 ай бұрын

there's a solution for this in the pinned comment

It would be great if you mention in video name or video itself (but better both) that this workflow is for SDXL.

@MDMZ

2 ай бұрын

the guide page has two workflows for 1.5 and sdxl

actually i have a i9 9900k and 16go ram, i dont have a graphics card. Do i need to get one ? i launch a comfyui video and its still not finish after 3hours. How quick is it with a graphics card ?

@MDMZ

3 ай бұрын

you definitely need one to run without issues, the speed depends on so many factors

Great video I got it to work on an m2 ultra 192gb. currently trying to integrate animatelcm and lcmlora to this workflow for faster generation and enhanced quality . let me know if you have any advice and ill stay tuned for more videos :)

@MDMZ

Ай бұрын

that's awesome, i'm really curious, how's the speed on the m2 ?

@murtazahashmi3690

28 күн бұрын

@@MDMZ its not super fast but it did get the job done. i ran it in 1 hour. on nvidia gpus i suppose it would be less than 30 mins for the same workflow

I am using 11GB gpu and gets an error when executing KSamplerAdvanced (Allocation on device 0 would exceed allowed memory. (out of memory)). Is there any workaround or my gpu is just too weak?

@kietzi

3 ай бұрын

same problem here.. just errors when its coming to the KSampler

@MDMZ

2 ай бұрын

This usually means u need a more powerful gpu with higher VRAM

@MDMZ

2 ай бұрын

What errors are u getting, did u check the pinned coment ?

@kietzi

2 ай бұрын

@@MDMZ yeah, I have just 8gb.. still waiting for my credit card to use the hosting services 😅 but I figured out, when I reduce the steps, it's working

@MDMZ

2 ай бұрын

@@kietzi glad to hear that

Somebody, plzz adviсe what should i do with this error " Error occurred when executing CheckpointLoaderSimple: 'model.diffusion_model.input_blocks.0.0.weight' "?

How do you adapt this flow the the newer comfyui animate diff loader ? "When loading the graph, the following node types were not found: AV_IPAdapter Nodes that have failed to load will show as red on the graph."

@MDMZ

9 күн бұрын

Hi, make sure to use the updated workflow (it's in the description)

When using open pose and hed - does that lock you into not changing the style or look of a character? You way seems more creative friendly in design.

@MDMZ

3 ай бұрын

hmm I'm not sure, I haven't tested that

@Nibot2023

3 ай бұрын

@@MDMZ *Edit solved it. I did a git pull on the ipadatper for an update and I made a ipadapter folder in the comfyui/models area and it worked. Awesome Tutorial - Going back throw and following along but for some reason I have the Ip adaptor in the same spot but for some reason the node is undefined. What would be the work around for that node to load the ip-adapter-plus_sdxl_vit-h.safetensors?

@MDMZ

3 ай бұрын

@@Nibot2023 Nice! thanks for sharing how you solved it

ksampler is not running for me before that the que gets stopped also i can't see any preview video how to fix this ???

@MDMZ

Ай бұрын

could be a memory issue, check the pinned comment

Thanks for the step by step tutorial. I am almost there... Can't figure out the below error tho: Error occurred when executing MiDaS-NormalMapPreprocessor: name 'midas' is not defined

@MDMZ

2 ай бұрын

can you share your workflow on discord ?

Work this with 3d render cutscenes too? I want shadow of rome looks better

@MDMZ

2 ай бұрын

I don;t see why not, you can try

Hi @MDMZ... Great video... I am running into an error tho that states. "RuntimeError: mixed dtype (CPU): expect parameter to have scalar type of Float" Do you know any solution as it would be really helpful. This happens at ksampler Thanks

@MDMZ

Ай бұрын

I don't think I came across this one yet, I'm adding solved errors to the pinned comment

Hi, i keep getting failed to fetch error tab, what does that mean? how can i fix it?

Great guide, my PC is fairly decent 3060 GPU and it takes forever to make a video, anything to speed it up? Ty

@MDMZ

3 ай бұрын

update your GPU driver, make sure your GPU is not being overtaken by other software

@biggo1261

Ай бұрын

@@MDMZ We can do with 3060 if so why not COLAB T4!! what about 3050TI 4GB RAM

Hi @MDMZ I'm having problems with "When loading the graph, the following node types were not found: MiDaS-DepthMapPreprocessor Nodes that have failed to load will show as red on the graph." can you help please or somebody?

@MDMZ

25 күн бұрын

which node is triggering this error ? can you share more info on discord ?

Thank you. no workflow json link?

@kietzi

3 ай бұрын

its in the guide the second file. he shows in the video. maybe watch again ;)

Awesome video 🤩

For me it works but sadly temporal coherence appears to be very bad. Do you have any idea how to fix that? Which numbers could I tweak

@MDMZ

Ай бұрын

there are several factors so it's all about experimenting, there's not a single setting that you can change to directly control that

KSampler stays at 33% Although I waited for 4.5 hours, it still did not work at the same steps 30, I tried at 25, it is the same again, the last time I was able to run it at 9, it also stayed at 33% Is there a solution? System: Ryzen 5 3600/gtx1070Ti 8GB/16GB 3200mhz Ram/500gb SSD

@MDMZ

3 ай бұрын

8GB might be a little low for it, but it could also be happening for another reason, did you try setting a lower resolution ? maybe 4080p

Hi, I'm not able to see the models in the checkpoint? I've them downloaded in the same location as done in the video.

@MDMZ

25 күн бұрын

hit refresh, try restarting otherwise

Where can I add Lora block? Thanks in advance.

My graphics card is 4070ti 12GB, and it takes 1.5 hours to generate a 6s long video. Is this normal?

@MDMZ

3 ай бұрын

a bit too slow, what resolution are you generating at ? maybe you can try to lower it

Hi! When loading the graph, the following node types were not found: IPAdapterApply Nodes that have failed to load will show as red on the graph. Any hints?

@MDMZ

16 сағат бұрын

Yes, use the fix workflow instead, it's in the description