Animatediff Comfyui Tutorial (How to Use Detailer) for Better Animation

[@zerox9646]

great workflow

[@goshniiAI]

I appreciative it. Thank you

[@97BuckeyeGuy]

I love how clean your workflow is. Well done.

[@goshniiAI]

i appreciate you for mentioning, and saying thank you.

[@mugglepower]

Really appreciate the tutorial brother. been struggling with ip adapter and animatediff for months. this immediately improved the results and plugs right into to my original workflow. subbed.

[@goshniiAI]

Welcome to the channel! I am glad the tutorial helped your workflow and improved your results with IP Adapter and AnimatedIFF. Thank you for your awesome feedback!

[@mugglepower]

@@goshniiAI is there a place where I could find some explanation for all the parameters on the face detailer nodes? also ipadapter?

[@goshniiAI]

@@mugglepower Yes for more detailed explanations, you can check out the channel of (Dr.Lt.Data's ComfyUI for face the detailer nodes) and (Latent Vision for Ipadapter)

[@colepeterson5392]

I REALLY appreciate how you built the workflow from scratch. You know your way around the nodes unlike a lot of people. Learned some tricks from this vid thanks

[@goshniiAI]

I am glad the process was easy to follow and practical, and pleased to hear you learned some new tricks along with you. Thank you for expressing your thoughts.

[@skybeast2738]

Great video and workflow, you're one of the best SD youtubers out there

[@goshniiAI]

I sincerely appreciate your feedback. Thank you so much, it really means a lot.

[@hbedrix]

amazing workflow and video, great job seriously

[@goshniiAI]

thank you for the motivating compliments. glad to hear from you.

[@PixelsVerwisselaar]

nice and thank you again for a great tutorial🙌. Question: How you got that vram and cpu etc ussage % thing in the right corner?

[@goshniiAI]

You're so welcome! Continue to experiment and create amazing things! The Resource Monitor installation instructions can be found here. kzbin.info/www/bejne/nITan3hog7B4sM0si=ZWWCCKjH3j_KdqCC

[@PixelsVerwisselaar]

@@goshniiAI I will sir 💪 and thank you again for the info 🫡💚

[@masoud.art.videos]

Thanks for clear explanation

[@goshniiAI]

you are most welcome.

[@joshuax2591]

Thanks a lot! But how to use face and hand detectors at the same time?

[@goshniiAI]

To use face and hand detectors at once, you will have to run two separate models or detectors.

[@ValleStutz]

Thanks for the tutorial!!! Is it possible to combine it with Controlnet and use the detailer with masks?

[@goshniiAI]

You're very welcome, and it is indeed possible to combine. Controlnet + Detailer with masks for better control over your animations.

[@ValleStutz]

@@goshniiAI does the controlnet have to be included for the detailer in the pipe?

[@goshniiAI]

@@ValleStutz Yes, it is necessary for controlnet to function properly. (POSITIVE AND NEGATIVE PROMPT) INTO (Apply controller) then return (to pipe positive and negative). Since this was not the exact lesson, give it a try and see if you gain the results

[@ValleStutz]

@@goshniiAI when i do it this way the Detailer puts a new person into the face

[@goshniiAI]

@@ValleStutz Alright, I'll have to give it a shot using ControlNet on my end.

[@elifmiami]

Its really nice workflow but when i tried did not worked somehow. My character a way bit smaller

[@goshniiAI]

Hello there, Thanks for the feedback! consider changing the input settings. Check that the initial settings are scaled correctly before applying the animation.

[@sybexstudio]

hello, i did everything and i get prompt output failed validation. Any input?

[@goshniiAI]

Check your prompt formatting, Sometimes, invalid characters or missing parameters can trigger this error. you can try simplifying your prompt.

[@sybexstudio]

@ I got it to work. Thanks! Another issue, the results u get always convert to an object in the center. Or it places a female character. What if I want just a landscape?

[@goshniiAI]

​@@sybexstudio⁣ You must use the Empty latent image node to change the aspect ratio. Any 16x9 ratio can be used, for example, width 896, height 512. you can scroll down here to find more aspect ratio sizes for SD1.5 www.winxdvd.com/ai-tips/best-image-size-for-stable-diffusion.htm

[@sybexstudio]

@ I didn't mean change aspect ratio. I mean I am getting weird results. CAn you explain in your workflow 2Controlnet +2upscale the Prompt you used? Why is it in this format. "0" :"Abstract watercolor painting of a woman, surrounded by (floral patterns:1.3) Background, surrounded by flowers, 8k, stunning intricate details, dynamic Scene by artgerm"

[@sybexstudio]

Also the Text node next to this.

[@Samosamossan]

Thank you for this great video! I just wanted to ask you something. I have tried to implement your workflow but with Video2Video instead of Text2Video, and the animatediff process works perfectly, but then the face detailer is completely messed up. Basically it pasts the new more detailed face but much smaller than the original face. As if the frame with the new face is in 256px, while the original frame is in 1024px. Do you have an idea on how to fix it? Thanks!

[@goshniiAI]

It sounds like you're having scaling problems with the face detailer. Double-checking the settings and parameters in the face detailer may help solve the problem. Additionally, you might try to ensure that all of your frames have the same resolution before sending them through the face detailer.

[@Samosamossan]

@@goshniiAI Thank you for your answer! I tried that, but it's not working. I tried different workflows and it works perfectly for text2video but there is always a problem with image2video, so I guess the problem comes from there. I will keep looking

[@goshniiAI]

@@Samosamossan I appreciate the effort you put into identifying and fixing! Sometimes the solution requires some experimentation and patience.

[@Deminii]

can u tell us how turn on this gpc cpu hdd info in ur queue window? I also see in others video that, solo controlnet also improve faces without adetailer, but adetailer i thinks is better

[@goshniiAI]

Thank you for the feedback, and you can find the resource monitor guide here: kzbin.info/www/bejne/nITan3hog7B4sM0si=H4DgxFGzpN-MoeDt

[@mattm7319]

workflow is great if I could only get the shutterstock logo off the bottom lmao i used cyberrealistic and a bunch of negative prompts and it still does it

[@mattm7319]

btw with upscaling this workflow is still super super fast on a 4090

[@mattm7319]

update: it was the animatediff model the sd_15 had probs, but v3 working fine

[@goshniiAI]

Thanks for the laugh and your comment. LOL , glad to hear about the solution with V3 Some models might handle stock watermarks better than others. negative prompts like "watermark" and "logo." Sometimes adding more specificity helps.

[@Injaznito1]

I have a question. I have a workflow that has these elements in this order with my model flow: checkpoint to IPAdapter group, IPAdapter group to lora group, lora group to animatediff group, animatediff to FreeU and from FreeU, finally to the Ksampler. my question is where do I place this group between? Does it go between my animatediff group and freeu? I also have my ControlNET group feeding the pos/neg prompt into the Ksamper so I'm guessing those will go to the ToBasicPipe instead of the Ksampler? Thanx!

[@goshniiAI]

Your workflow appears to be quite complex. As you thought I would recommend inserting your group between the Animatediff group and FreeU. Regarding the ControlNET group, you are correct; the pos/neg prompts should be routed to the ToBasicPipe rather than the Ksampler. I hope this hepls.

[@Injaznito1]

@@goshniiAI Well, I thank people like you who post extremely helpful information :). I'm spanking new (1 month) to Comfyui but have been able to piece together these modules by following you guy's tutorials. Thank you, thank you, thank you!

[@goshniiAI]

@@Injaznito1 You are most welcome, and I'm glad you found the videos useful. Thank you for the feedback.

[@jonathanbeaton6984]

Hey! Did you ever figure this out? I am trying the same thing and having the same issue!

[@SonidosEnArmonia_1992]

why are you using the mm_sd_v14 model? isnt that older? or what

[@goshniiAI]

You are correct, the mm_sd_v14 model was selected based on a personal choice. With future projects, I hope to test out the new Animatediff (model V3)

[@Bicyclesidewalk]

Gumroad is not working...

[@goshniiAI]

Hello there, you can also find the link here: goshnii.gumroad.com/

[@Bicyclesidewalk]

@@goshniiAI Ok - it is working now. Cool, thanks - will try it out later tonight.

[@goshniiAI]

@@Bicyclesidewalk Awesome!!

[@zeta.luxurylifes]

💫thanks a lot, keep carry on. use some ai for your sound (make more quality).

[@goshniiAI]

I appreciate your suggestion, Thank you Lots!