pettarin/forced-alignment-tools

A collection of links and notes on forced alignment tools

forced-alignment-tools's Issues

Quality/performance of the alignment tools

How is the is the quality of the alignements of the different methods? Or how would one even assess the performance of the produced alignments?

Include my hosted Replicate API endpoint for Aeneas

Hey! Great work on this repo, it helped me to get up to speed quickly on what's out there.

I've deployed a public Aeneas endpoint if you don't wanna mess around with Aeneas installation and deployment.

Can I add it to the Readme?

Here's the link, the interface and Replicate docs are self-explanatory https://replicate.com/andreemic/forced-alignment

Maximum length?

Dear Alberto,

thanks a lot for this great list! Even though you do not list this property in the table, I would be interested to hear if you have any insights as to the maximum length of recordings and transcripts for the various tools. At the Distributed Little Red Hen Lab, we need to align one-hour recordings with their transcripts. This works great with Gentle for English, but for instance the Montreal Forced Aligner is only happy with relatively short passages. I'd be grateful for any pointers you have.

Best regards,
Peter

Suggestion: Include granularity

As stated in the README:

A text fragment can have arbitrary granularity:
a paragraph,
a sentence,
a portion of a sentence (i.e., a group of words),
a word, or
a phoneme (i.e., a single sound).

This information would be useful to include in the table.

While aligners such as Gentle and SPPAS allow phone-level alignment, others such as aeneas can only perform word-level alignment.

Perhaps there could be a column indicating the granularity of each aligner?

Did you implement some of these?

hello ,Did you implement some of these?
I tried to implement some of these, but not one running successfully.
Thanks

Which forced alignment tool can be run on Windows?

I have an audio and its transcript. And I'd like to get a timestamp for each word of my transcript, but not a new list of words with timestamps.
Which one do you think would be better?