Face Tracker

Face Tracker 0.7.0

Supported Bit Versions
  1. 64-bit
Source Code URL
Minimum OBS Studio Version
Supported Platforms
  1. Windows
  2. Mac OS X
  3. Linux
This plugin provide a filter for face detection and face tracking for mainly a speaking person. Major applications are a sermon at a church, a presentation at a conference or company's meeting but not limited.

This plugin employs a library called dlib to detect a face and track it.
The frame of the source is periodically taken to face detection algorithm. Once a face is found, the face is tracked.
Based on the location and the size of the face under tracking, the frame will be cropped and expanded.

A simple demo on OBS Studio (no audio)​

Limitations and Known Issues
  • Currently the plugin works for one face. If there are two or more faces, the plugin might hover back and forth over the faces.
  • Consumes a lot of CPU and memory resource. It is recommended to have enough test beforehand.
  • If zoom is vibrating due to inaccuracy of face detection or tracking, please consider to set 'scale max' property and set 'zoom' to the maximum. This will fix the zoom by 'scale max' property.
  • MacOS and Ubuntu binaries are not tested.
  • Accuracy of the result is not guaranteed.
You can donate to me via PayPal, GitHub, or Ko-fi.
First release
Last update
4.70 star(s) 10 ratings

More resources from norihiro

Latest updates

  1. Face Tracker 0.7.0

    This is an experimental release. Changes in 0.7.0 Implemented a dock widget to monitor and...
  2. Face Tracker 0.6.4

    This is a hotfix change release. Changes in 0.6.4 Linux: Fixed the plugin is not loaded by...
  3. Face Tracker 0.6.3

    This is a hotfix change release. Changes in 0.6.3 Fixed an issue the filter is not loaded by...

Latest reviews

Super cool! Wanders a little much.

Sometimes gets up close in your face spontaneously.

However, for a method of realigning your camera and pausing it it is incredible!

Hope the Wandering gets fixed. then it will be near perfect functionality
This is a great start, but a bit unstable and unpredictable.
It is better than the Nvidia thingy!

>are you able to add option to track not a face, but specific colour?
>For example a surfer's colour shirt? or RC car colour?

I think what would be very cool would be to get a general object classification system into obs.

So anything in the feed would be tagged in a way that other plugins could use.

Object 111, Person A, T-Shirt, Bounding Box 123,123,255,255,...
Object 222, Person B, Head, Bounding Box 123,123,255,255,...

Person A's head -> Plugin B could use it to attach a text object showing brand of t-shirt

Person B's head -> Plugin C could use it to "fuzz it out" for anonymization

etc. etc.

I think demand for something like a Nvidia GPU is totally acceptable for something like this as well. Don't need to be CPU only.

I am not in this business for commercial reason, but just from personal interest, I would be happy to sponsor work here with some small amount of money. Anyone else? Is there a way to sponsor OBS plugins?
are you able to add option to track not a face, but specific colour?
For example a surfer's colour shirt? or RC car colour?
Just a simple colour picker (with some margin) instead of Landmark detection data
Good options are offered and personalization are given while leaving the whole customization accessible and user friendly.

Would include more landmark detection data/model
Works just as advertised!
Excellent plugin! wanted to ask if there's a way to update this with the ability to add a source that could "follow" your head.
For example, a hat that will always be in top of your head
I love this plugin specially with the integrated PTZ control.... Probably another good feature that can be added to this is the ability to switch to another scene when a detected face reach a certain region in the frame.
Excellent work to date, this is a critical part to having an automated camera person for Houses of Worship and other venues.
This plugin has much potential. It feels a bit like the new feature introduced by Apple on the M1 ipads called "center stage".

It does seem to slowdown my laptop, but the developer does mention the (potential) high load on CPU and memory so I guess it is fair.

I am grateful for the developers work and I hope he (and with help from the community) makes this plugin even better! It would be amazing if it could consume less resources in the future.