Most online caption tools send your video to a server so it can be transcribed there. For an interview, a client call or a video that is not public yet, that is a risk you can avoid. A caption tool that runs in your browser does the same job and the file stays on your device.
The speech recognition model has to run on your computer. That means a one-time model download (about 41 MB for the small model) and some waiting: on a slow laptop a few minutes of audio can take about as long as the audio lasts. After the first download it works without an internet connection.
Names, numbers, product terms and anything said over music are the usual mistakes. Clear speech with little background noise gives the best result.
Open the free Caption MakerLonger files and offline launch files are planned for a one-time $24 pack, not yet for sale: Pack in preparation