EchoMimic: Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning

https://github.com/antgroup/echomimic

EchoMimic Series

• EchoMimicV1: Lifelike Audio-Driven Portrait Animations through Editable Landmark Conditioning. GitHub
• EchoMimicV2: Towards Striking, Simplified, and Semi-Body Human Animation. GitHub
• 
  

For V1 you just need a Headshot - the software will find and crop the head automatically for you

For V2 you need a 1/2 body shot, there are 20 female and 20 male in the Input directory already - and also in the custom node if you dig deep enough and look for a directory called - "ref_imgs_by_FLUX".

Testing and Costing

Ultra - EchoMimic

Bargain 99 cents an hour

V1 model (face only)

First run - 243 seconds 6.7 cents

Second run - 174 seconds 4.8 cents

V2 model (half body)

First run 225 seconds 6.2 cents

Second run 176 seconds 4.8 cents