Request - incrementaly saving the generated audio chunks

[credo99]

Hi, I am back with a feedback related to the Colab CPU runtimes (as opposed to the GPU ones). The time required for audio generation is huge compated to GPU and also the audio has many segments of mumbo-jumbo that renders it useless. I tried the GPU runtimes until I ran intro a temporary GPU ban losing also a considerable amount of work at the end of a generation cycle when Google forcibly disconnected my runtime. I had 240 characters chunking in a text of about 3000 characters so I lost almost 10 chunks that were already generated when I was disconnected. I would be nice if the engine and the GUI would have a boolean parameter called something like "save audio chunks" (in native wav format since I assume that the mp3 conversion is done on the full set) so when clicked it would save incrementally the generated chunks in a folder on the colab or on the gdrive (mounted previously) with a incremental name derivd from the generation timestamp (up to miliseconds).This would allow recovering the already generated chunks in case of a session / runtime errot or disconnect. Thanks a lot