Live Portrait - How to make talking AI Avatars in ComfyUI?

Make Talking Avatar style videos in your local PC using ComfyUI for FREE!
In this tutorial we'll learn how to generare an avatar in any custom style using your own face. Then how you can animate that using a simple video input. All in two simple comfyUI workflows.
⬇️Download the workflow here:
drive.google.com/drive/folder...
The json metadata is embeded in the png files. Simply drag and drop them to your ComfyUI browser window and load the workflow.
Install the missing custom nodes using ComfyUI manager.
🔗Links for the custom nodes:
LivePortrait: github.com/kijai/ComfyUI-Live...
InsightFace: github.com/Gourieff/Assets/bl...
IP Adapter: github.com/cubiq/ComfyUI_IPAd...
Instant ID: github.com/cubiq/ComfyUI_Inst...
SDXL TurboVision: civitai.com/models/215418/tur...
SDXL Emoji LoRA: civitai.com/models/144245/sdx...
#liveportrait #talkingavatar #aitalkingvideo #aivideo #aitutorial #kazitorials #aitalker #ailipsync #aitalkingvideo
Chapters:
00:00 Intro
00:30 How to create an AI avatar from your own photo
01:40 How to create talking AI avatar from your own video
02:45 How to install Insightface properly
04:42 Outro

[@labmike3d]

Hi Kazi, I'm thrilled to hear that you're enjoying LivePortrait! I also did a quick test with Hedra retargeting and was very pleased with the results. It's amazing to see how well these tools work together. Keep having fun and exploring new possibilities! 😊

[@kaziahmed]

Thanks :) yeah hedra is awesome!

[@CalsProductions]

Cool tutorial thanks not many give you clear instructions, you did and it worked a treat!

[@kaziahmed]

Thank you 😊 I am glad it worked for you 🙌🏽🙌🏽

[@hippoevolutionsociety8436]

hi kazi, i dont think it works anymore since the update. crop info error. can you please redo with lip sync ? thanks

[@kaziahmed]

Oh, I will check and update, thanks!

[@PigeonShrine]

my "load instantid Model" node isn't working, i have the same files as yourself but doesn't read what it needs to from the bin file. could you attach new file links for me to download so i can try this?

[@kaziahmed]

The main model can be downloaded from HuggingFace and should be placed into the ComfyUI/models/instantid directory. huggingface.co/InstantX/InstantID/resolve/main/ip-adapter.bin?download=true

[@garau]

Hi kazi. Can u recomend us which kind of pc do we need for using comfyui? Which vcard? Nvida ?thx

[@kaziahmed]

Hey great question! For getting started with comfyUI I'd recommend get a decent pc with a dedicated GPU that has at least 12GB of VRam, so nVidia 3080 and above would be great. For the rest of the pc parts, get that's appropriate for that GPU. Check PC Part Picker website for more recommendations.

[@amaru_zeas]

Great video thanks :). I am getting this though: Error occurred when executing LivePortraitProcess: PyTorch is not linked with support for mps devices do you know how to fix it?

[@kaziahmed]

What GPU are you using? This might not work on anything other than nvidia gpus.

[@amaru_zeas]

@@kaziahmed RTX 4090

[@kaziahmed]

@@amaru_zeas I will do some digging and reply to your query. Not sure why that happened. Do you have insightface set up properly? Does the Reactor node or instantID work on your comfyUI environment?

[@dollarproduction24]

I cannot find the code while installing the Insightface ID on the Comfy UI terminal. Can you please provide the code here?

[@kaziahmed]

.\python_embeded\python.exe -m pip install insightface-downloaded-file-link onnxruntime

[@dollarproduction24]

@@kaziahmed Thank you so much ❤❤

[@mj88leong]

how to we just output the Avatar instead of combine?

[@kaziahmed]

These are two different workflows. The first one is for just avatar and the second one is for live portrait. Just watch the demonstration in the video :)

[@kaischaffer5188]

Can you upgrade the load image node to also accept a video input?

[@kaziahmed]

For the first workflow (which is for avatar) or the second one (for live portrait)?

[@kaischaffer5188]

@@kaziahmed for the LivePortrait workflow :)

[@kaziahmed]

@@kaischaffer5188 The live portrait imports videos, that's how it works. Did you download my workflow? See the demonstration in the video please :)

[@kaischaffer5188]

@@kaziahmedwhat I referred to was the image load (source image). How can I upload a source video that will be animated by another driving video? Hope that you understood what I meant. Thanks :) And great video

[@kaziahmed]

@@kaischaffer5188 oh… umm. Nout sure if that’d work. Check the live portrait github page for more details on this.

[@Hotich.1996]

could we bring it on live stream ?

[@kaziahmed]

Not yet, it's not real time. But hopefully soon!

[@user-qz2sh4ss8f]

what 's gpu card??

[@kaziahmed]

GPU = graphics processing unit, basically your graphics card.

[@emmaqias7176]

woooow

[@kaziahmed]

Try it out and share your results in the comments!

[@emmaqias7176]

​@@kaziahmedI will do it tomorrow.