Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

VLC tops 6B downloads, previews AI-generated subtitles

techcrunch.com

Post reply on HN
Print view View original

81–90 of 163 posts

Re: VLC tops 6B downloads, previews AI-generated subtitles

#81

Post by lxgr »

eviks wrote: post #66
jl6 wrote: post #52

There's an art to subtitling that goes beyond mere speech-to-text processing. Sometimes it's better to paraphrase dialog to reduce the amount of text that needs to be read. Sometimes you need to name a voice as unknown, to avoid spoilers. Sometimes the positioning on the screen matters. I hope the model can be made to understand all this.

> better to paraphrase dialog to reduce the amount of text that needs to be read. That's just bad destructive art, especially for a foreign language that you partially know. > Sometimes you need to name a voice as unknown, to avoid spoilers. Don't name any, that's what your own eye-ear voice recognition/matching and positioning are for (also reduces the amount of text) > Sometimes the positioning on the screen matter…

> Don't name any, that's what your own eye-ear voice recognition/matching and positioning are for

That’s tricky when one or more speakers aren’t visible.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#82

Post by ttoinou »

jillesvangurp wrote: post #64
n144q wrote: post #30

me: waiting years for VLC to fix basic usability issues and reconcile UI across different platforms VLC: we're gonna work on AI

This boils down to software development not being free. In VLC's case, the development is funded by several for profit companies (like videolabs) that make their money from stuff they do with VLC (consulting, commercial services, etc.). VLC is a good example of an OSS project that is pretty well run with decades of history that has a healthy ecosystem of people and companies earning their living supporting all that a…

They don’t need designers when they are the free media player that stood the test of time and being used by the masses. It’s true organic bottom up design, tweaked little by little over the years

Re: VLC tops 6B downloads, previews AI-generated subtitles

#83

Post by lidavidm »

jl6 wrote: post #52

There's an art to subtitling that goes beyond mere speech-to-text processing. Sometimes it's better to paraphrase dialog to reduce the amount of text that needs to be read. Sometimes you need to name a voice as unknown, to avoid spoilers. Sometimes the positioning on the screen matters. I hope the model can be made to understand all this.

I suppose this feature should have been termed closed captioning and not subtitling. It seems you're not going to get much sympathy for human translation here.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#84

Post by _flux »

CaptainFever wrote: post #70
_flux wrote: post #63

Earlier quoted context omitted.

What is the alternative? The results could be cached, but it's probably unlikely that they would need to be used again later, as I imagine most videos are watched only once. Another option would be to upload the generated subtitles to some service or p2p, but I believe that would also problem (e.g. privacy, who runs the service, firewalls for p2p, etc).

OpenSubtitles as a cache, maybe? (With their collaboration, of course.)

It seems like some kind of review process would need to be included to reduce abuse possibilities.

But yes, that would be quite nice, if there are enough people who don't mind uploading the file names of the media they play (or some other unique media identifier?) along with the subtitles to that service—and with user credentials? It would certainly need to be opt-in, which makes one wonder if it would be much effective at all.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#85

Post by nicoburns »

daemin wrote: post #42

Isn't this what YouTube has been doing with automatically generated subtitles for videos for years? I've always found their quality to be somewhat lacking.

Quality definitely can be lacking, on the other hand, it is significantly better than no subtitles at all.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#86

Post by Hard_Space »

Aurornis wrote: post #77
jl6 wrote: post #52

There's an art to subtitling that goes beyond mere speech-to-text processing. Sometimes it's better to paraphrase dialog to reduce the amount of text that needs to be read. Sometimes you need to name a voice as unknown, to avoid spoilers. Sometimes the positioning on the screen matters. I hope the model can be made to understand all this.

AI subtitles are just text representation of the sound track. There is no need for artistic interpretation, substituting words, or hiding information. If it’s in the audio, there’s no reason to keep it out of the subtitle. An AI subtitle generator that takes artistic license with the conversion is not what anyone wants.

That doesn't work for idioms, certainly in Italian, which has multiple colorful metaphors which would be mystifying if translated directly.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#87

Post by _flux »

rvnx wrote: post #75
_flux wrote: post #63

Earlier quoted context omitted.

What is the alternative? The results could be cached, but it's probably unlikely that they would need to be used again later, as I imagine most videos are watched only once. Another option would be to upload the generated subtitles to some service or p2p, but I believe that would also problem (e.g. privacy, who runs the service, firewalls for p2p, etc).

It can actually send to Google the information on what you are playing: https://github.com/videolan/vlc/blob/f908ef4981c93a8b76805ad... and to their own servers: https://github.com/videolan/vlc/blob/f908ef4981c93a8b76805ad... should could fetch subtitles as the same time ? edit: cf, what "a3w" says too.

Downloading is still much easier to handle than uploading.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#88

Post by Dr4kn »

torben-friis wrote: post #19

I recently used some subtitles that I later found out had been AI generated. The experience wasn't really good to be honest - the text was technically correct but the way multiline phrases were split made them somehow extremely hard to parse. The only reason I learned AI was involved is that it was bad enough for me to stop viewing and check what was wrong. Hopefully it's an implementation detail and we'll see better…

Proper subtitles are obviously better, but it's impossible to do on everything. The tech is going to get better, and is already a game changer for hearing impaired people. Subtitles that are mostly correct are much better than none at all.

VLC has the option to find subtitles. If you use plex or jellyfin there are add-ons or bazarr, which does it automatically

Re: VLC tops 6B downloads, previews AI-generated subtitles

#89

Post by scarface_74 »

_flux wrote: post #63
c16 wrote: post #47

Very excited for this, but a waste of energy if everyone is needing to process their video in real time.

What is the alternative? The results could be cached, but it's probably unlikely that they would need to be used again later, as I imagine most videos are watched only once. Another option would be to upload the generated subtitles to some service or p2p, but I believe that would also problem (e.g. privacy, who runs the service, firewalls for p2p, etc).

Tangentially related: funny enough, this site was just submitted yesterday to HN - https://exampl.page/

You can navigate to $foo.exampl.page and it will generate a website on the fly with text and graphics using AI. It will then save and cache the page.

It’s admittedly a useless but cool little demo.

Re: VLC tops 6B downloads, previews AI-generated subtitles

#90

Post by deneas »

compootr wrote: post #6
cynicalsecurity wrote: post #4

Please no AI.

> using open-source AI models that run locally on users’ devices, eliminating the need for internet connectivity or cloud services this isn't the bad kind of ai we hate... it could be actually useful!

Unfortunately the term "open-source" AI alone is meaningless without knowing the users definition of open-source. More often than not it just means "openly available". In this case for example they could be using Whisper, which is openly available, but from OpenAI and still has all the caveats around used data, resource usage for training etc.
Post reply on HN