For a while now, I've been really struggling with the playback controls. The playback of a narrative, whether previously generated or whether it's just regenerating and playing, is not at all smooth, and I think it needs an overhaul. There are several issues that contribute to this: Firstly, Hume creates the generation sections at random places, often halfway through a paragraph or sentence. So it can stop and start generating halfway through something it's saying, which can prevent it flowing well and produce some weird voice variations halfway through a sentence. When I press play at the bottom, I have the "Until End" toggle turned on. So I expect it to keep playing until it reaches the end of the whole text. This happens only in one situation: when all the text HASN'T been generated. So pressing play at the start of an ungenerated piece of text, will use up credits until it reaches the end of the text. That's not unexpected in itself, BUT: If any text has been generated previously, it pauses the playback at that text. So if I have generated some parts of the text, when I play it back, it keeps stopping when it reaches a pre-generated part. This is hugely annoying, because I can't listen to a whole chunk of text cleanly to make sure it all flows. This then leads onto another issue. When it hits a pre-generated piece of text, it pauses playback, but it visually selects the next piece of text it's already generated. So I hit play, thinking it will play the next part. However, instead of playing the next part, it replays the text it just played to me, so it loops back essentially, instead of carrying on playing. so to play a whole chunk of text just to check it sounds right, it will keep pausing and interrupting and I have to keep faffing around with it, by which time I've forgotten how the previous part sounded. A further issue exacerbates this. When I select a piece of text and add "Acting Instructions", it treats that part as a separate generation. So if I add acting instructions to just one word inside a paragraph, I can't play the whole paragraph cleanly. The playback stops at the acting instructions, and I have to manually select the next part and press play again. Changing the playing selection to "Selection" does nothing. It just plays by generated section as normal, and doesn't play the selected text. ______________________ My suggestion: The voice generation sections should follow the paragraphs, this seems intuitive and obvious. So the generation of a chunk won't stop halfway through a sentence any more. Obviously if a really long paragraph it could be split, but only at a new sentence. Playback should just work. If I have "Until End" toggled on, playback should just keep playing until the end of the text, regardless of what's been generated, and regardless of the sections / chunks it was generated in. If I have "Selection" toggled on, the playback should just play that selection. Seems much simpler right? Am I missing something?
