About the audio tools
Audio formats split into lossy and lossless. MP3, AAC in an M4A, and Opus in an OGG throw away detail your ears are unlikely to miss and land around one megabyte a minute. WAV keeps every sample uncompressed at roughly ten megabytes a minute, and FLAC compresses that to about half with nothing lost. Converting lossless to lossy is a one-way trip, and converting one lossy format to another compounds the loss, so go back to the original whenever you have it.
The same FFmpeg build that handles video does the audio here, so extracting a soundtrack, trimming a podcast, merging takes, or normalizing volume all run on your own machine. Speech to text is different: it runs Whisper on your graphics card where WebGPU is available, and the transcript never leaves the tab either.