BradKML
commented
3 years ago Checking on previous issues, someone noted https://github.com/carykh/jumpcutter/issues/123 https://github.com/carykh/jumpcutter/issues/44
Open HaujetZhao opened 4 years ago
BradKML
commented
3 years ago Checking on previous issues, someone noted https://github.com/carykh/jumpcutter/issues/123 https://github.com/carykh/jumpcutter/issues/44
To CaryKH:
Thank god I can meet your JumpCutter, really astonished me and simplified my video cutting process! Thank you so much! I believe the idea of JumpCutter will be legendary in the history of Video Editing.
And after watching your video demonstrating the JumpCutter, I also got puzzled of how to recognize the gesture and then autocut the footage.
After days of thinking, searching, practicing, implementing, a brilliant idea came to me, it doesn't require any, any AI technology!
Let's go deep into the essential goal you want to achieve. Facially, you want to autocut the video by recognizing the hand gesture, which is a marker. But, your essential goal is, to add a marker into the video while shooting, and let the program auto-cut by recognizing these markers.
That's the key! Adding a marker and recognizing a marker!
Let's see the solution:
Back to yesterday, I wanted to add a subtitle to my video, I'm lazy, so naturally, I choose letting the commercial Speech-To-Text service, uploading the audio, the subtitle is generated perfectly, the timestamp are exact!
Today, it suddenly came to me, why we have to recognize the gesture, if we can recognize the voice?
And we even don't need to recognize the voice, because tons of cheap service already can auto generate the subtitles!
So here is the idea: while we are shooting, when we are content with the last clip, just say a key sentence like "save" , if we want to discard the last clip, just say a key sentence like "cut it".
After the shooting, first we use JumpCutter to remove the silence clips with a margin of 10 frames. Then we upload the audio to get a auto-generated subtitle. ( Many services can recognize our key words perfectly ). Then, we will use the JumpCutter again to delete the dissatisfied clips according to the timestamp by searching the key word in the subtitle. And this time, we can give it a smaller margin, like 5 frames. Then, a perfect video is made! ( by the way, the key word you set must be easy to recognize, and only search the sentence with only the key word, so that would not injure the normal sentence, regex can do this search)
So my expression is done, Than you for your reading!
The idea is not complex, but since I am even not a amateur developer, I can't implement the simple idea into python code. So I share it to you, wishing it can be implemented.
Also, I submitted a merge pull to your jumpcutter, hoping you can agree it. It mainly resolved two problems:
\1. If the last process was interrupted, the TEMP folder won't be removed, when next time the jumpcutter runs, it would went error. So I added a code to detect if the TEMP folder exists and delete it before creating it.
\2. When the file path includes spaces, the ffmpeg command would went wrong, I fixed it.
Again, Thank you sincerely! Your jumpcutter really made my day!
Hoping to receive your opinions soon!
also @kool601 @pniedzwiedzinski @Ethelt