This is great! I would love an explicit "set bookmark to here" button if I scroll up from the high watermark

I discovered the reset, but it took a minute and I expect given the way I read being able to explicitly walk the high watermark back a bit would be nice (maybe that's doable if you select text? I didn't try!)

The format is nice, I love standard ebooks

I opened Moby Dick but didn't see how to listen to it on my phone, not sure what the status of the "listen" effort is.

I listened to some books from standard ebooks using eleven labs, but I'm really excited about a better implementation

▲

samuelcole 8 hours ago | parent | next [-]

just did the explicit 'read from here' line, to set the watermark precisely (it also controls where the play button starts from)

check it out! let me know what you think

▲

samuelcole 2 days ago | parent | prev [-]

moby dick didn't align properly, there's a confidence gate it didn't pass.

i know, it makes me sad too, i'll keep working on it

this is the list of everything: https://tale.fyi/audiobooks

librivox is all human volunteers, so i think it makes a difference from eleven labs, but i might add a fallback to ai readings for folks who want that

reset feedback noted!

▲

p-e-w 2 days ago | parent [-]

> moby dick didn't align properly, there's a confidence gate it didn't pass.

What does this mean?

▲

samuelcole 2 days ago | parent [-]

well! (i've been waiting for someone to ask ha)

when i first knew i wanted to align audio to text, i reached for aeneas, which was built for this exact problem/product (https://www.readbeyond.it/aeneas/)

however, it drifted by hundreds of seconds deep into chapters and emits no confidence signal, so i couldn't eval it cleanly

so i switched to a neural acoustic model trained with ctc (meta's massively multilingual speech aligner, it ships inside torchaudio): it outputs a probability for each character at every frame of audio, and viterbi decoding finds the most probable path that spells exactly our text, with a per-word probability

that probability is the gate: a book ships read-along only if its median confidence and coverage clear a bar. moby dick's didn't, so it's honestly text-only

▲

p-e-w a day ago | parent [-]

Could the problem be Moby Dick’s infamously large vocabulary, which contains words used essentially nowhere else?

▲

samuelcole a day ago | parent [-]

i actually had an out-of-memory bug, which i've since fixed. chewing through it now to see if we can get it

	▲	samuelcole a day ago \| parent [-]
		got it working! https://tale.fyi/moby-dick it got confused by some of the more uhh complex passages ;-)