Skip to content
SPECIFICTION
Get listed

Robust Video Matting

github.com

Free plan WebLinuxiOSAndroidAPISelf-hosted
Visit github.com
Spec score 8.2 #13 of 29 in AI Video Background Removers

What it is

Robust Video Matting is a free project for separating people from video backgrounds. It processes frames with a recurrent neural network that carries temporal memory, and the project says it can perform matting in real time without additional inputs. A browser webcam demo runs the model live and can visualize its recurrent states. The conversion API accepts a video or image sequence and can produce a composited video or PNG sequence, optionally including alpha and foreground outputs.

Models are available for PyTorch, TorchHub, TorchScript, ONNX, TensorFlow, TensorFlow.js, and CoreML. MobileNetv3 is recommended for most uses; ResNet50 is larger and described as offering small performance improvements. The README reports GPU throughput figures, including 4K at 76 FPS and HD at 104 FPS on a GTX 1080 Ti, and 4K at 154 FPS and HD at 172 FPS on an RTX 3090 with FP16 under specified settings. The project cautions that file conversion runs much slower because its script does not use hardware video encoding or decoding or parallel tensor transfers. CoreML models require iOS 13 or later and do not support dynamic resolution. The code is GPL-3.0 licensed.

Spec sheet

Purpose
Robust Video Matting is designed for human video matting and uses a recurrent neural network with temporal memory to process video frames.Read 9 Oct 2026 · github.com
License
The repository states that the code is released under the GPL-3.0 license.Read 9 Oct 2026 · github.com
Browser demo
The webcam demo runs the model live in a browser and can visualize recurrent states.Read 9 Oct 2026 · github.com
Colab demo
The Colab demo lets users test the model on their own videos with a free GPU.Read 9 Oct 2026 · github.com
Compatibility limits
CoreML does not support dynamic resolution, and the provided ONNX models use opset 12 and are tested with CPU and CUDA backends.Read 9 Oct 2026 · github.com
Examples and training
The repository provides a Colab demo for testing videos with a free GPU and links to training and evaluation documentation.Read 9 Oct 2026 · github.com
Frameworks
The project provides model implementations or exports for PyTorch, TorchScript, ONNX, TensorFlow, TensorFlow.js, and CoreML.Read 9 Oct 2026 · github.com
iOS requirement
The provided CoreML models require iOS 13 or later, and CoreML does not support dynamic resolution.Read 9 Oct 2026 · github.com
Mobile support
CoreML models require iOS 13 or later, and the README links a third-party NCNN C++ Android project.Read 9 Oct 2026 · github.com
Model choices
The project recommends MobileNetv3 for most use cases and describes ResNet50 as a larger variant with small performance improvements.Read 9 Oct 2026 · github.com
Model formats
The project provides models for PyTorch, TorchHub, TorchScript, ONNX, TensorFlow, TensorFlow.js, and CoreML.Read 9 Oct 2026 · github.com
Model variants
MobileNetv3 is recommended for most use cases, while ResNet50 is larger with small performance improvements.Read 9 Oct 2026 · github.com
No additional inputs
The project says it can perform matting in real time on videos without additional inputs.Read 9 Oct 2026 · github.com
Performance
The README reports 172 FPS for HD and 154 FPS for 4K on an RTX 3090 using FP16, with batch size 1 and frame chunk 1.Read 9 Oct 2026 · github.com
Performance limitation
The project cautions that its video conversion script is expected to run much slower than the reported tensor throughput because it does not use hardware video encoding or decoding or parallel tensor transfers.Read 9 Oct 2026 · github.com
Project origin
The repository says the project was developed at ByteDance Inc.Read 9 Oct 2026 · github.com
Real-time performance
The README reports 4K at 76 FPS and HD at 104 FPS on an Nvidia GTX 1080 Ti GPU.Read 9 Oct 2026 · github.com
Real-time processing
The project says it can perform matting in real time without additional inputs.Read 9 Oct 2026 · github.com
Support and guidance
The README links inference and training documentation for usage, training, and evaluation instructions.Read 9 Oct 2026 · github.com
Video conversion
The provided conversion API accepts a video file or image sequence and can output a composited video or PNG sequence, with optional alpha and foreground outputs.Read 9 Oct 2026 · github.com

Plans and prices

Robust Video Matting Free GPL-3.0 · pretrained models and source code Read 9 Oct 2026 · github.com

Every Robust Video Matting plan and price →

Strong points and limits

Strong points

  • Real-time matting without additional inputs
  • Browser webcam demo can show recurrent states
  • Conversion accepts video files or image sequences
  • Optional alpha and foreground outputs
  • Models span seven listed formats

Limits

  • File conversion runs much slower than tensor throughput
  • CoreML requires iOS 13 or later
  • CoreML does not support dynamic resolution
  • Code is licensed under GPL-3.0

Who it suits

This project suits developers who need human video matting and want model options across machine learning frameworks or a live browser demo. iOS users need iOS 13 or later, and CoreML users must account for its lack of dynamic resolution support.

Verdict

Robust Video Matting offers several model formats, a browser demo, and file conversion with optional outputs. The reported GPU figures describe tensor throughput, so they do not represent the slower conversion script.

Where Specifiction places it

Alternatives to Robust Video MattingAll alternatives →