NowDigital HumanIt’s really popular! No need to appear on camera or record, just use ChatGPT to write the text and you can directly produce the video.

But the digital people available on the market now are either very expensive or impossible to register. I think it would be great if there was a local free version!

I found this, today I brought you a game called SadTalker Digital human generation software, which can be run independently locally, and has been made into an integrated package that can be run with one click.

1. System requirements

Operating system: Windows 10 and above

Memory: 16G

CPU: I5 or above

Graphics card: NVIDIA 30 or 40 series or above, 8G video memory

Hard disk: 20G or more

Software download address: Link: https://pan.baidu.com/s/1B_lcgZrKqsMqwcq5bWu7rg?pwd=i83f

2. SadTalker drives digital human in pictures

1) Double-click "SadTalker_WEBUI Run.bat", press ctrl and click the link to enter the web page

2) Upload the image and audio file on the web page and click the "Generate" button

For specific parameters, please refer to the figure below:

3) Generate a digital person of your photo

The generated digital human video is in the results directory.

3. SadTalker_Video drives the video digital human

a) Record your own video

It is best to record a half-body video. The body and hands can move, but the mouth should not move. The head can turn and move appropriately, but the movements should not be too big. The face should face the camera, and special attention should be paid to not moving the mouth.

2) Use Jianying to adjust the video and audio to the same length

Here we use Jianying to generate a digital person with a still mouth as an example. The method is to let the digital person read: Hmmmmmmmm…

3) Save video files and audio files separately

Special note: Please use a full-digit file name, otherwise there is a certain probability of error.

4) Double-click "Drive Video to Generate Digital Human.exe" and select the "Voice Drive Video Digital Human" page

Select parameters as shown below.

5) Generate your own video digital human

The generated digital human video is in the results_bat directory.

IV. Conclusion

Using the offline version of SadTalker to create a digital human has relatively high system requirements. It is recommended to process the video in segments, with each segment not exceeding 60 seconds.

In the AI era, the central bank has begun to use digital human technology. Everyone can be a host, even a newbie with no basic knowledge. As long as you want, AI can help you monetize your side job.