1 # Facial Landmarks for Cubism
3 A library that extracts facial landmarks from a webcam feed and converts them
4 into Live2D® Cubism SDK parameters.
6 *Disclaimer: This library is designed for use with the Live2D® Cubism SDK.
7 It is not part of the SDK itself, and is not affiliated in any way with Live2D
8 Inc. The Live2D® Cubism SDK belongs solely to Live2D Inc. You will need to
9 agree to Live2D Inc.'s license agreements to use the Live2D® Cubism SDK.*
11 This block diagram shows the intended usage of this library:
13 ![Block diagram showing interaction of this library with other components](block_diagram.png)
15 Video showing me using the example program:
16 <https://youtu.be/SZPEKwEqbdI>
18 ## Spin-off: Mouse Tracking for Cubism
20 An alternative version using mouse cursor tracking and audio based lip
21 syncing instead of face tracking is available at
22 <https://github.com/adrianiainlam/mouse-tracker-for-cubism>.
24 The main advantage is a much lower CPU load.
26 ## Supporting environments
28 This library was developed and tested only on Ubuntu 18.04 using GCC 7.5.0.
29 However I don't think I've used anything that prevents it from being
30 cross-platform compatible -- it should still work as long as you have a
31 recent C/C++ compiler. The library should only require C++11. The Cubism
32 SDK requires C++14. I have made use of one C++17 library (`<filesystem>`)
33 in the example program, but it should be straightforward to change this
34 if you don't have C++17 support.
36 I have provided some shell scripts for convenience when building. In an
37 environment without a `/bin/sh` shell you may have to run the commands
38 manually. Hereafter, all build instructions will assume a Linux environment
39 where a shell is available.
41 If your CPU does not support AVX instructions you may want to edit "build.sh"
42 and "example/demo.patch" to remove the `-D USE_AVX_INSTRUCTIONS=1` variable
43 (or change AVX to SSE4 or SSE2). However there could be a penalty in
48 1. Install dependencies.
50 You will require a recent C/C++ compiler, `make`, `patch`, CMake >= 3.16,
51 and the OpenCV library (I'm using version 4.3.0). To compile the example
52 program you will also require the OpenGL library (and its dev headers)
53 among other libraries required for the example program. The libraries I
54 had to install (this list may not be exhaustive) are:
56 libgl1-mesa-dev libxrandr-dev libxinerama-dev libxcursor-dev libxi-dev libglu1-mesa-dev
58 2. Clone this repository including its submodule (dlib)
60 git clone --recurse-submodules https://github.com/adrianiainlam/facial-landmarks-for-cubism.git
62 3. To build the library only: (Skip this step if you want to build the example
63 program. It will be done automatically.)
65 cd <path of the git repo>
68 4. You will require a facial landmark dataset to use with dlib. I have
70 <http://dlib.net/files/shape_predictor_68_face_landmarks.dat.bz2>.
71 Extract the file and edit the "config.txt" file to point to the
74 Note: The license for this dataset excludes commercial use. If you want
75 to use this library in a commercial product you will need to obtain a
76 dataset in some other way.
78 To build the example program:
80 5. Copy the extracted dlib dataset from step 4 to the "example" folder
83 6. Download "Cubism 4 SDK for Native R1" from the Live2D website:
84 <https://www.live2d.com/en/download/cubism-sdk/download-native/>.
86 Extract the archive -- put the "CubismSdkForNative-4-r.1" folder under
87 the "example" folder of this repo.
89 Note: The Cubism SDK is the property of Live2D and is not part of this
90 project. You must agree to Live2D's license agreements to use it.
93 "example/CubismSdkForNative-4-r.1/Samples/OpenGL/thirdParty/scripts"
98 8. Go back to the "example" directory and run
102 9. Now try running the example program. From the "example" directory:
104 cd ./demo_build/build/make_gcc/bin/Demo/
108 ## Command-line arguments for the example program
110 Most command-line arguments are to control the Cubism side of the program.
111 Only one argument (`--config`) is used to specify the configuration file
112 for the Facial Landmarks for Cubism library.
114 * `--window-width`, `-W`: Specify the window width
115 * `--window-height`, `-H`: Specify the window height
116 * `--window-title`, `-t`: Specify the window title
117 * `--root-dir`, `-d`: The directory at which the "Resources" folder will
118 be found. This is where the model data will be located.
119 * `--scale-factor`, `-f`: How the model should be scaled
120 * `--translate-x`, `-x`: Horizontal translation of the model within the
122 * `--translate-y`, `-y`: Vertical translation of the model within the window
123 * `--model`, `-m`: Name of the model to be used. This must be located inside
124 the "Resources" folder.
125 * `--config`, `-c`: Path to the configuration file for the Facial Landmarks
126 for Cubism library. See below for more details.
129 ## Configuration file
131 Due to the differences in hardware and differences in each person's face,
132 I have decided to make pretty much every parameter tweakable. The file
133 "config.txt" lists and documents all parameters and their default values.
134 You can change the values there and pass it to the example program using
135 the `-c` argument. If using the library directly, the path to this file
136 should be passed to the constructor (or pass an empty string to use
141 1. Example program crashes with SIGILL (Illegal instruction).
143 Your CPU probably doesn't support AVX instructions which is used by dlib.
144 You can confirm this by running
146 grep avx /proc/cpuinfo
148 If this is the case, try to find out if your CPU supports SSE4 or SSE2,
149 then edit "build.sh" and "example/demo.patch" to change
150 `USE_AVX_INSTRUCTIONS=1` to `USE_SSE4_INSTRUCTIONS=1` or
151 `USE_SSE2_INSTRUCTIONS=1`.
155 The library itself is provided under the MIT license. By "the library itself"
156 I refer to the following files that I have provided under this repo:
158 * src/facial_landmark_detector.cpp
160 * include/facial_landmark_detector.h
161 * and if you decide to build the binary for the library, the resulting
162 binary file (typically build/libFacialLandmarksForCubism.a)
164 The license text can be found in LICENSE-MIT.txt, and also at the top of
165 the .cpp and .h files.
167 The library makes use of the dlib library, provided here as a Git
168 submodule, which is used under the Boost Software License, version 1.0.
169 The full license text can be found under lib/dlib/dlib/LICENSE.txt.
171 The example program is a patched version of the sample program provided
172 by Live2D (because there's really no point in reinventing the wheel),
173 and as such, as per the licensing restrictions by Live2D, is still the
176 The patch file (example/demo.patch) contains lines showing additions by
177 me, as well as deleted lines and unchanged lines for context. The deleted
178 and unchanged lines are obviously still owned by Live2D. For my additions,
179 where substantial enough for me to claim ownership, I release them under
180 the Do What the Fuck You Want to Public License, version 2. The full license
181 text can be found in LICENSE-WTFPL.txt.
183 All other files not mentioned above that I have provided in this repo
184 (i.e. not downloaded and placed here by you), *excluding* the two license
185 documents and files generated by Git, are also released under the Do What
186 the Fuck You Want to Public License, version 2, whose full license text
187 can be found in LICENSE-WTFPL.txt.
189 In order to use example program, or in any other way use this library
190 with the Live2D® Cubism SDK, you must agree to the license by Live2D Inc.
191 Their licenses can be found here:
192 <https://www.live2d.com/en/download/cubism-sdk/download-native/>.
194 The library requires a facial landmark dataset, and the one provided by
195 dlib (which is derived from a dataset owned by Imperial College London)
196 has been used in development. The license for this dataset excludes
197 commercial use. You must obtain an alternative dataset if you wish to
198 use this library commercially.
200 This is not a license requirement, but if you find my library useful,
201 I'd love to hear from you! Send me an email at spam(at)adrianiainlam.tk --
202 replacing "spam" with the name of this repo :).
206 Contributions welcome! This is only a hobby weekend project so I don't
207 really have many environments / faces to test it on. Feel free to submit
208 issues or pull requests on GitHub, or send questions or patches to me
209 (see my email address above) if you prefer email. Thanks :)