Nothing BS about it. In ML, end-to-end means feeding raw data (e.g. text) to the model and getting raw data (e.g. waveform audio) out. This is on contrast to a

Nothing BS about it. In ML, end-to-end means feeding raw data (e.g. text) to the model and getting raw data (e.g. waveform audio) out. This is on contrast to approaches that involve pre- and postprocessing (e.g. sending pronunciation tokens to the model, or models returning FFT packets or TTS parameters instead of raw waveforms). It's a common and well understood technical term in this context.