Congratulations, I haven't used it on stream yet, but I've already installed it and carried out a simple test. It seemed very accurate in terms of real-time translations. I'm going to recommend it to some friends who live on Twitch like me. Thank you very much for sharing your work with us.
I installed this new version and what's happening is that the last sentence of the translation is always visible, I changed several settings but it didn't help, can you help me? Thank You.
this video was good and i got my system to get real time captions locally. I initially i was having problems but after making sure the right Cuda and Obs updated versions were all integrated correctly i was good to go. thanks again for this video.
Cool repo! I'd like to have captions in Spanish when I speak English, and captions in English when I speak Spanish, so I don't neglect any of my viewers from different countries. Is this possible with LocalVocal? Thank you.
can we get a docker based http server type of solution? I'd love to use it through my raspberry pi and just stream the audio (ideally through ws/wss) and get the response back from the server.... (btw tried your solution, but defender immediately started screaming about it :D ). If you want to I can collaborate with you on this project, I just can't take this alone, because I have no idea on how the OBS side of things have to be implemented.
Hi Roy Shilkrot, I hope this message finds you well. I am writing to provide some feedback on the voice recognition plugin you developed. I have encountered a few issues and have some suggestions that I believe could enhance the functionality of the plugin: Support for Simplified Chinese: Currently, the plugin only supports Traditional Chinese. It would be greatly beneficial if it could also support Simplified Chinese, as it is widely used in many regions. Language Setting Reset: When specifying Chinese as the language, if the speech contains any other language, the setting automatically resets to auto-detect. It would be more user-friendly if the specified language setting could remain fixed unless manually changed by the user. GPU Usage: Even when the plugin is hidden in the source box, it continues to run and occupy GPU resources. It would be helpful if hiding the plugin could also pause its operation to save on processing power. Thank you for your time and consideration of these suggestions. I look forward to any updates that may address these issues. Best regards
Some of these issues I know and I'm fixing. Some others I may need more help You could open issues on the GitHub repo, or meet us in the server discord.gg/CJHr5zHXD3 for logs sharing etc.
I'm seeing quite a lag. There are gaps in my understanding but.. I've been wondering if there is a way to tighten it up by sending the CC as a separate load rather than baking it in to the stream video. That way the encoding time that OBS is adding would serve to offset the time that whisper takes to generate the CC. This of course means that the player in the browser at the end is laying the CC over the video. But I think that's the better way of doing CC anyway??
i've added the option for "partial" captions and now there is virtually no lag at all, it's almost instantaneous, except you do need a decent GPU to handle the real-time partial transcription.
There are "filters" you can apply, look for the button in the advanced settings But the best way is to add Noise Cancellation / Suppression filter before (above) L.V. This reduce the problem by whole lot
@@royshilk Hi thx for your reply.. before I texted you yesterday.. I already add the suppression filter for my setting's standard ... so can you give me another option to disable that...?
@royshilk how to disable the sencetes like "thank you for watching", like and subscribe,...etc... that sentences automatically appear while translating...
I wonder what's the reason It might be possible but I'd like to understand what about the built in feature makes it easier or better for you so I can build it in localvocal as well Thanks
I want to install this version of ChatGPT. Or should I install the local one? My computer has an RTX 3050 Ti laptop. It seems like if I do it via the GPT API, my computer will be less strained. However, I don't know how to configure the GPT settings in the new local version. The options shown in the old video are no longer available.
as amd user i can tell it is useless since it freezes, lags and works veeeeery slow but thanks for your hardwork! maybe one day it will be done also for amd users!
Great vid!! i have a question though. I've been trying to install this (the windows 64 cuda installer). Unlike the video, I can't specify my download location (just goes straight into extracting) and even though the installer claims to have successfully install the plugin, I can't see the Local Vocal audio filter option once I restart OBS. any ideas about what I missed and/or can do?
if localvocal isn't showing up there's a compatibility issue either with OBS version or drivers etc. please meet us on discord.gg/J5RgzMmPqM for live support
@@royshilk I am using software (not CUDA), witch is impracticable on my Ryzen 4600g, even the tiniest model, freezes my machine. But the model only is effectives after restarting OBS, and it startups with an "loading model bar" screen on OBS's startup.
@@royshilk that's what I did though, but as stated in my own reply, the problem was that it didn't download the models and I have to restart OBS. But now translation is a bit scuffed, I have to put base whisper translation for FR to EN, the other translation models are not working properly (nllb is completely fucked, doesn't translate then translate in other languages at random and add random symbols, M2M doesn't make sense most of the time and MADLAD is just not willing to work) For EN->FR I select M2M 1.2B, it works well (nllb still does weird stuff (doesn't translate adds a male symbol at the end of every sentences, translate twice with different translations and sometimes put the phonetics, and MADLAD doesn't translate)
@@royshilk by the way, what's the directory location for downloaded models? I'd like to delete the useless ones Also don't think I don't appreciate your work, I really do, we need more people like you doing open source free stuff on the net, thank you
@@Silails this is where the models are saved "C:\Program Files\obs-studio\data\obs-plugins\obs-localvocal\models" you can just delete the folders you don't need