ComfyUI - Complete AI Influencer Workflow

Hi everyone!

So recently, after seeing Plastic Bowl's post on BHW, I have fallen down the rabbit hole of AI Influencers, specifically creating a consistent model within SD with ComfyUI, and I've become obsessed.

I've gone pretty balls deep on learning and researching how it all works and how to use it, but being a complete beginner to anything like this, it is taking a while to learn - although I am pleased with the rate of improvement I am making.

My question is; does anybody have a ComfyUI workflow that they use to execute the majority of the influencer creation needs? For example, a full-kit workflow that, based on prompts, will create a scene, whilst injecting a ImageLoad defined face ,and body type (for that consistency).

Essentially, I am looking for a complete workflow that someone is willing to share, or suggest one they have found online which they now use.

I am looking forward to learning all this through trial and error along the way, as of course that's how to learn best, however I would, alongside learning, like to be able to actively progress with building the following for an AI influencer as well, which a working workflow today would be the last piece of the puzzle - a bit like putting together all the outer edges first and considering it complete lol.

Any advice or guidance would be greatly appreciated! If anyone knows of any useful Discord chats or similar where likeminded people communicate that would also be lovely!

Looking forward to discussing!

Looking for the same, if u guys have one, share please!!

Can you share links of some youtube videos to understand the basics of it

Use new gen flux which is far better than SD and SDXL or even par with midjourney and open source. You can run locally. For consist face just train a lora with like 10-50 images.

I have complete all in one workflow for ai influencer even for sdxl even without lora just using ipadapter faceid + reactor for faceswap. Just dm me if you want.

Don't know much about Comfy workflow and its capability, but why not use models hosted on hugging face, replicate you can even mantain consistency and use faceswap for AI influencer or Fooocus for creating the influencer .

Or is there something I don't know like what exactly does comfy can do that these can't ir it being locally hosted the selling point, as we do have to add .safetensors from different models???

MugOfSteel said:

Don't know much about Comfy workflow and its capability, but why not use models hosted on hugging face, replicate you can even mantain consistency and use faceswap for AI influencer or Fooocus for creating the influencer .

Or is there something I don't know like what exactly does comfy can do that these can't ir it being locally hosted the selling point, as we do have to add .safetensors from different models???

Click to expand...

Using ComfyUI give lot of control to you. Those spaces on huggingface is good for quick task, it's not meant for quality work.

While fooocus seems good but lacks lot of functions like control net, IP adapters. It only have quick faceswap which isn't good.

ComfyUI is like A1111 it's webui to do AI related stuffs (mainly image generation and video generation) it's locally and you have to load safetensors files in it.

rexlapix said:

Using ComfyUI give lot of control to you. Those spaces on huggingface is good for quick task, it's not meant for quality work.

While fooocus seems good but lacks lot of functions like control net, IP adapters. It only have quick faceswap which isn't good.

ComfyUI is like A1111 it's webui to do AI related stuffs (mainly image generation and video generation) it's locally and you have to load safetensors files in it.

Click to expand...

Okay I think I get it but as you said you have "I have complete all in one workflow for ai influencer even for sdxl even without lora just using ipadapter faceid + reactor for faceswap. Just dm me if you want." could you share it with me?

Im looking for the same. I want a workflow that creates consistent content like this every run. It has a load image node that i can apply a headshot so it can faceswap or deepfake.
@katya.quinnx / instagram

Im looking for just images not video just yet. Any specialist here, hmu.