The core function of artificial intelligence is to “reduce costs and increase efficiency”. Without this premise, talking about how advanced and awesome a certain AI tool is, is definitely just playing with concepts.

When it comes to video editing, many people may think that it is a job that only professionals can do, and that it requires a lot of effort to learn various editing software.

Imagine if your boss assigns you an important task and asks you to cut the highlights or key parts of a long video into a short video, what would you do?

For example, your boss participated in an interview on a program. After receiving the master tape, he asked you to edit out the part where he spoke in the interview separately. What would you do?

Is the only way to do this is to work overtime to review the entire video at 1.5 times the speed, and then edit it frame by frame? Is there a more efficient way? It’s best not to work overtime. Today I made an appointment with my card-playing buddies to play mahjong after get off work.

It doesn’t matter. There is an open source and free AI video editing tool now. It can automatically edit videos with one click and help you complete the above work in 5 minutes.

FunClip is an open source AI editing tool produced by Alibaba, a domestic Internet giant. It uses AI technology (the open source FunASR Paraformer series models of Alibaba Tongyi Voice Laboratory) to accurately recognize speech in videos. Based on the text in the recognition results, users can quickly select the required text/speaker and crop it into a video clip.

FunClip features:

Automated speech recognition

FunClip integrates Alibaba’s industrial-grade model Paraformer-Large, which is a leader in speech recognition with high accuracy and precise prediction timestamps. This allows users to quickly find specific content in the video through speech recognition.

Hot word customization

Sometimes, there are some specific words in the video that we pay special attention to, such as a person’s name or a specific event. FunClip allows users to specify these hot words through the integrated SeACo-Paraformer model to improve the recognition accuracy of these words.

Speaker Recognition

FunClip integrates the CAM++ speaker recognition model. This feature allows users to crop video segments of specific speakers based on automatically identified speaker IDs. This is very useful for video clips that need to distinguish different speakers.

Video Cropping

Users can select a text segment in the recognition result or specify a speaker and click the crop button to obtain the corresponding video segment. This feature makes video editing simple. You no longer need to manually drag the timeline, saving a lot of time.

Multi-clip support

FunClip also supports users to edit videos in multiple segments, providing flexible editing capabilities. This means that users can edit videos more carefully according to their needs.

FunClip can be deployed locally, that is, downloaded to the computer and configured with dependent environments, so that it can be used permanently and for free indefinitely, even without an Internet connection. Of course, if you don’t know how to download and install open source code from GitHub, you can also visit the following website for a free experience.

Github project address: https://link.zhihu.com/?target=https%3A//github.com/modelscope/FunClip

Magic Tower Experience Website:

https://modelscope.cn/studios/iic/funasr_app_clipvideo/summary

HuggingFace experience website:

https://link.zhihu.com/?target=https%3A//huggingface.co/spaces/R1ckShi/FunClip

It is very simple to operate.

Step 1: Upload your video

The second step is to distinguish the speakers (if there are multiple speakers in the video) and set hot words. This step is based on personal needs, and you can choose not to choose.

The third step is to extract and recognize the language in the video and convert it into text

In the fourth step, you can copy paragraphs from the text extracted in the previous step to the “text to be cropped” or enter the “speaker to be cropped” so that the AI ​​knows which paragraph to crop from.

The fifth step is to set subtitle parameters. This step is not required.

Step 6: Edit the video

The final edited video is generated very quickly, almost in seconds.