Then there are four main stages of the program: But this stage is constant (does not depend on the length of the video) and takes about 30 seconds. Image by Authorįirst, you need to read the vosk model. Program execution time for a 17-minute video. The first word indicates the beginning of the fragment to be cut, and the second - its end. In Russian, they are written as начало and конец, and pronounced like nachalo and konets, respectively. The text will be recognized as nonsense, but we don’t need it, right? The main thing is that the control words will be recognized correctly. Since vosk has foreign models, you can recognize English speech using your native language model. I decided to just use the words start and end in my native language. Because I’m not a native English speaker, I decided to move on to a simpler solution. Also at this moment, I realized how much the accent affects speech recognition quality. But it turned out that rare words are worse recognized, and very rare ones may not be in the dictionary at all. Initially, I decided to use some rare English words like sagacious and egregious. And if you say, the program will work incorrectly. In fact, of course, you can, but you can’t be sure that you won’t say these words in your usual speech. We need two - for the start and the end of the cut fragment. That’s great! Now we need to choose control words - there is room for creativity. We know the time of pronouncing each word. Automatic ( completely automatic, without human intervention) - identifying long moments of silence.
But I want to make a fast, simple, and free program. For example, wisecat, which can do much more than “video cutting”. It usually takes more than three hours to simply edit an hour-long video! That’s why I came up with this idea. You need to review the video completely (and possibly more than once!), select all fragments you need, join them and then render the video for a long time. Removing unnecessary video fragments is not a difficult task, but a long one.
0 kommentar(er)
