Personally, the examples and explanation was too minimal at times and I had a pretty difficult time understanding the majority of it, I had to re-watch many times to get it since I am not a programmer, it's not well explained for your average joe... And in the end so if I am understanding, this isn't for speech translation to subtitles? Should state it in the title...
I'm trying to translate Korean into English, but I get an error like this. Can you tell me the solution? -> ERROR message : Your request contained invalid JSON : 'utf-8' codec can't decode byte 0xa9 in position 98 invalid start byte
Thanks for trying the plugin! Yes there's probably a utf8 encoding problem to figure out. Can you plz open an issue? github.com/obs-ai/obs-polyglot/issues I can track and resolve the problem
Excellent plugin. I really appreciate your work. Would it be possible to set a language as excluded from translation such as English and have the plugin detect and translate any language spoken or shown that is non-English? For example, if a streamer's native language is Danish but primarily streams speaking English, is it possible to have the Danish words show up as English captions? This is the type of implementation that I believe would be incredibly valuable.
I believe the model is able to auto detect the language. But I'm not sure about mid-sentence changes or mix of languages, that's advanced. I'm pretty sure though that OpenAI / ChatGPT can do it, if you explain that in the prompt
Polyglot is really easy to install and setup thanks to your tutorial. Question, what is the language code for Japanese? I can't seem to find the language code in the Github readme for the life of me.
Thanks for plugins! I am facing a problem that local polyglot server is not able to correctly translate text if it has a newline symbol. It maybe won’t be an issue but ocr plugin adds this symbol during text capture which leads to translation failure. Is there any possibility to fix this?
When I setup the data source on polyglot I don't have "text" on Dynamic input but I have none/facebook/twitter/instagram/youtube. Also when I push test request I get SSL connect error, can you help me?
the "text" in my video is because i have a Text Source on my OBS scene that I named "text". The dropdown there will scan all your scenes for Text Source and display their names so you choose one. You shouldn't get SSL errs unless you're going for "" instead of "". Try just "127.0.0.1:18080" and you should be Ok
hi, i just want to say it works with localvoice and thank you! but i'm a bit confused whyy when i no longer speaking(nothing to transcript) polyglot respond with i'm not sure. i'm translating from bahasa indonesia to english. my output parsing options are text, empty regex,empty regex flags, empty post process regex. and when i press test request the parsed output are i'm not sure. i tried fill the empties(regex,regex flags,post process regex) with whats in the video but the result are Failed to match regex. sorry i have 0 knowledge about this things, i just tried this because it seems cool and i want to try this for my streams. edit: i fiddle a bit, and i manage to get the parsed output into empty. by having empty regex, empty Regex flags, post process regex into : I'm not sure then replace with empty. although i think i just replace the text of i'm not sure into . but since i'm doing bahasa indonesia into english i think its fine for me. hope i can get to proper solutions but at least with the current state i'm pretty satisfied. again this is very cool stuff.
@@royshilk like i already stated, i pretty much copy paste based on image shown in your video, with a difference of : {"text":"{{input}}", "source_lang":"ind_Latn","target_lang":"eng_Latn"} Note: For people that tried it the first time capital L on Latn are important the very first time i tried without considering case sensitive it doesn't work. then what works for me to make test request working and the plugin to works are below: Output Parsing Options: Content-Type: Text Regex: Regex Flags: Regex Group=0 Post process Regex: I'm not sure Replace: again i have 0 knowledge about how it works or what Regex even mean, but when i tried to copy according to the image on the video when i press test request it says in red text above the button as : Failed to match regex. maybe i'm missing something or didnt read the documentation properly and i'm sorry for that, hopefully this comment somehow help with the development, cheers! tldr i think i basically just tell the AI to replace its phrase "I'm not sure" which is the idle response if there is no additional input from localvocal(blank input) into [blank] if i understand it correctly P.s: i tried english > indonesia too, and the idle response are berbagai(means several) and i did the same to it too(changing the AI response of blank input which is "berbagai" into [blank] or . . .
there could be many issues... like your windows OS version or how you installed OBS. can you share a link to your OBS logs? you can get the link from the Help menu in OBS
@@royshilk No AJA devices found, skipping loading AJA UI plugin 22:20:02.380: Failed to initialize module 'aja-output-ui.dll' 22:20:02.425: No AJA devices found, skipping loading AJA plugin 22:20:02.425: Failed to initialize module 'aja.dll' 22:20:02.427: Skipping module '../../obs-plugins/64bit/chrome_elf.dll', not an OBS plugin 22:20:02.562: [CoreAudio encoder]: Adding CoreAudio AAC encoder 22:20:02.677: Skipping module '../../obs-plugins/64bit/ctranslate2.dll', not an OBS plugin 22:20:02.681: Failed to load 'en-US' text for module: 'decklink-captions.dll' 22:20:02.687: Failed to load 'en-US' text for module: 'decklink-output-ui.dll' 22:20:02.708: A DeckLink iterator could not be created. The DeckLink drivers may not be installed 22:20:02.709: Failed to initialize module 'decklink.dll' 22:20:02.875: Skipping module '../../obs-plugins/64bit/libcef.dll', not an OBS plugin 22:20:02.876: Skipping module '../../obs-plugins/64bit/libEGL.dll', not an OBS plugin 22:20:02.877: Skipping module '../../obs-plugins/64bit/libGLESv2.dll', not an OBS plugin 22:20:03.154: Skipping module '../../obs-plugins/64bit/libopenblas.dll', not an OBS plugin 22:20:03.286: [obs-browser]: Version 2.22.1 22:20:03.286: [obs-browser]: CEF Version 103.0.5060.134 (runtime), 103.0.0-5060-shared-textures.2594+gc69ad37+chromium-103.0.5060.134 (compiled) 22:20:03.692: [noise suppress]: NVIDIA denoiser disabled, redistributable not found or could not be loaded. 22:20:03.693: Failed to get NVVideoEffects.dll version info size 22:20:03.693: [NVIDIA VIDEO FX]: FX disabled, redistributable not found or could not be loaded. 22:20:03.774: LoadLibrary failed for '../../obs-plugins/64bit/obs-polyglot.dll': The specified module could not be found. 22:20:03.774: (126) 22:20:03.774: Module '../../obs-plugins/64bit/obs-polyglot.dll' not loaded 22:20:03.956: [obs-websocket] [obs_module_load] you can haz websockets (Version: 5.3.3 | RPC Version: 1) 22:20:03.956: [obs-websocket] [obs_module_load] Qt version (compile-time): 6.5.3 | Qt version (run-time): 6.5.3 22:20:03.956: [obs-websocket] [obs_module_load] Linked ASIO Version: 102800 22:20:03.989: [obs-websocket] [obs_module_load] Module loaded. 22:20:04.078: [vlc-video]: Couldn't find VLC installation, VLC video source disabled 22:20:04.177: ---------------------------------
Can you please actually copy in the http link to the logs from OBS? There's an option to get a link to the logs and if you send it I will get the full picture of what's happening
Hi, once download the models what do i do next, sorry for the silly question, i am so new at this and i want to translate for a church serviced in Vietnamese to English's. @@royshilk
Hello, I tried to set up the plug-in, the first time no problem, but I can't find the language code for French (I'm from France) so I cross my fingers and enter : {"text":"{{input}}", "source_lang":"fr_Latn", "target_lang":"eng_Latn"} (French is not in the readme list) the translation is slow and does not include all words. and the second time I launch OBS, I can't get it to work, even the "translation failed" error message doesn't appear. But bravo for the work already accomplished
This is a list of all lang codes huggingface.co/facebook/nllb-200-distilled-600M/blob/main/special_tokens_map.json It's slow because it runs local on your computer If you're missing words make sure the input includes all the words
Hey, I was wondering where the language codes are for the body section? I checked the readme over a few times but cant seem to find them. Was also wondering if you can translate into other scripts besides latin? Thanks for the great plugin! :)
Here's a link where all language codes are described: github.com/facebookresearch/LASER/blob/main/nllb/README.md It can definitely translate to much more than just Latin scripts...
Sentence Piece Model. you can download from pretrained-nmt-models.s3.us-west-2.amazonaws.com/CTranslate2/nllb/flores200_sacrebleu_tokenizer_spm.model if you're using NLLB
I have the error after downloading and selecting "Failed to load CT2 model. Unable to open file 'model.bin' F:/*filepath*" I tried all three of the NLLB-200 multilingual model, but all of them gave me the same error. Any way you can help?
Did you point the settings to your models? Seems like the plugin simply isn't configured to find the model filed. Tomorrow I'm streaming "office hours" so you're welcome to join and ask your question and I'll help in real time!
Hello ! Thank you for your work. it seems that after uninstalling Polyglot on Windows, the language files stay on "AppData\Roaming\obs-studio\plugin_config\"
Right it's possible the uninstaller doesn't know about downloaded files and won't delete them.. not sure how to make it do that though. Would need to research
do you have a tutorial to install in windows? It looks very different on my end and I have no idea what I am doing. It isnt showing up things on my obs.
Thanks ! But it says for me "Failed to Load Model CT2 Model. File model.bin is incomplete: failed to read a buffer of size 262354944 at position 191" I really don't understand what does that even mean-
Did you download the spm and set up the file in the settings? The SPM is downloaded from pretrained-nmt-models.s3.us-west-2.amazonaws.com/CTranslate2/nllb/flores200_sacrebleu_tokenizer_spm.model
When using this plugin, my CPU utilization reaches 100% (Windows 11 user). The system specifications include an i5-11400H processor, 16GB RAM, and an RTX 3050 4GB graphics card. taskbar showing: obs (23) and chrome (20).
Yes the local translation is heavy on the CPU. It's running an LLM on your PC which will be costly. on your machine I would recommend running a cloud translation. You can use OpenAI chatgpt for licey translation or a service like DeepL, which you can integrate with my URL/API Source
this might be a noob question, but i can't seem to find a Data source in obs (i'm using windows version, i see you're using Mac version). Is your data source the same as Text (GDI+) on windows? or should I be using a Browser source instead?
I don't see anywhere in the read me the language codes. I get you're an engineer not a teacher, it's like a meme issue with engineers lol. But you should be more detailed with your walk throughs. Don't skip steps that are obvious to you. If you say something is somewhere, show it and how to get there. Not trying to be rude, just I'm not even used to obs yet, and it's the 3rd video I've seen of yours where I'm just unable to follow you. And I'm pretty tech savvy ....
Thanks for the feedback! I'm trying to get better, but at the same time I value your time and I won't make videos with fluff and baits. The codes are indeed mentioned in the readme but here they are github.com/facebookresearch/LASER/blob/main/nllb/README.md
@@royshilk Thank you for adding those, because they weren't there before. But also, no non latin languages? I don't see Korean, Japanese, Or Chinese or anything of the sort. Thought it was no language left behind lol
@@davidmorton6406 The list given only contains less used languages added with LASER3, it doesn't contain the codes for english, french, italian, etc. either. You can find a more complete list by going to the side bar, opening the "laser_encoders" folder, and looking in the "language_list.py" file.
you actually don't need JSON parsing - the output of polyglot server is clear text. so please just use "Text" for the parsing option (instead of "JSON")
@@royshilk yeah i use it but i have always this error messagge sometimes. I use cloud close captions and write captions in a text source. Then url source and setting all likes the video... If you want i can write to you on obs forum and send some screen if you need them
@@MannaggACardarellyes! I appreciate screenshot, also you can open an issue github.com/obs-ai/obs-polyglot/issues so I can track The plugin is still waiting approval on the forum..
Please try the release github.com/occ-ai/obs-urlsource/actions/runs/6758096048#artifacts Until I release the official v0.2.1 I'm just waiting on a final pull request valuation from another tester
@@royshilk How can I know what the abbreviation for Portuguese is: I would like to speak Portuguese and translate it into English. But I don't know what the abbreviations of the words would be.
@@royshilk Thanks I didn't pay attention to the model selection when I had first set it up. I ended up using MangaOCR to transcribe japanese characters to text in order to feed the model
hi how are you its very good plugin i am trying to install it but i do not know where i get the problem how i will take your help in this which communication source if you will be guide me i will be very thankful. after every thing its shows me the jason error
Does this output text that is "embedded" into the video stream, or is the user able to turn it on and off? And does it only translate to one language at a time?
you can output to a text source - in which case it would be "embedded", or you can select in URL/API Source to output the result to the current stream as captions - in which case your viewers can turn on / off right now it's single language, but if you use something like OpenAI LLM API you can do multiple language at once, just ask the GPT that
@@royshilk I have a dutch friend who wants to show real time english subtitled translation for english viewers. Maybe this is already possible? but i'm struggling to achieve that at the moment. Using your guide but not seeming to get it working.
I'm happy to help get the translation up and running You can find me on the OBS discord server, or DM on X/Patreon etc. Also I'm live streaming and helping people with plugin problems twice a week TuTh, you can join that
Sir, you're a life saver... You should get a Nobel prize.... You should get two just in case you loose one.....
Personally, the examples and explanation was too minimal at times and I had a pretty difficult time understanding the majority of it, I had to re-watch many times to get it since I am not a programmer, it's not well explained for your average joe... And in the end so if I am understanding, this isn't for speech translation to subtitles? Should state it in the title...
Thank you for the feedback! I'm learning how to best make titles and descriptions such that people know what to expect.
The video is awesome, can't wait to see how to do it for realtime caption translation ! 😁
Coming soon! I'll update
I'm looking forward to this as well! Well done!
I'm trying to translate Korean into English, but I get an error like this. Can you tell me the solution? -> ERROR message : Your request contained invalid JSON : 'utf-8' codec can't decode byte 0xa9 in position 98 invalid start byte
Thanks for trying the plugin! Yes there's probably a utf8 encoding problem to figure out. Can you plz open an issue? github.com/obs-ai/obs-polyglot/issues I can track and resolve the problem
Excellent plugin. I really appreciate your work. Would it be possible to set a language as excluded from translation such as English and have the plugin detect and translate any language spoken or shown that is non-English? For example, if a streamer's native language is Danish but primarily streams speaking English, is it possible to have the Danish words show up as English captions? This is the type of implementation that I believe would be incredibly valuable.
I believe the model is able to auto detect the language. But I'm not sure about mid-sentence changes or mix of languages, that's advanced. I'm pretty sure though that OpenAI / ChatGPT can do it, if you explain that in the prompt
@@royshilk Very cool thank you
Polyglot is really easy to install and setup thanks to your tutorial. Question, what is the language code for Japanese? I can't seem to find the language code in the Github readme for the life of me.
the lang codes are in github.com/facebookresearch/LASER/blob/main/nllb/README.md (this link is mentioned in the polyglot readme)
hello awesome inovation; it's future. thanks.
currently the plugin have a built-in download model option in the dock so we don't have to download the model file
Thanks for plugins!
I am facing a problem that local polyglot server is not able to correctly translate text if it has a newline symbol. It maybe won’t be an issue but ocr plugin adds this symbol during text capture which leads to translation failure.
Is there any possibility to fix this?
Yes easy, add a header "Content-Type" "application/json"
That'll add escaping
When I setup the data source on polyglot I don't have "text" on Dynamic input but I have none/facebook/twitter/instagram/youtube. Also when I push test request I get SSL connect error, can you help me?
the "text" in my video is because i have a Text Source on my OBS scene that I named "text". The dropdown there will scan all your scenes for Text Source and display their names so you choose one. You shouldn't get SSL errs unless you're going for "" instead of "". Try just "127.0.0.1:18080" and you should be Ok
Pretty cool but I couldnt find portuguese on the list
hi, i just want to say it works with localvoice and thank you!
but i'm a bit confused whyy when i no longer speaking(nothing to transcript) polyglot respond with i'm not sure.
i'm translating from bahasa indonesia to english.
my output parsing options are
text,
empty regex,empty regex flags, empty post process regex.
and when i press test request the parsed output are i'm not sure.
i tried fill the empties(regex,regex flags,post process regex) with whats in the video but the result are Failed to match regex.
sorry i have 0 knowledge about this things, i just tried this because it seems cool and i want to try this for my streams.
edit: i fiddle a bit, and i manage to get the parsed output into empty.
by having empty regex, empty Regex flags, post process regex into : I'm not sure
then replace with empty.
although i think i just replace the text of i'm not sure into .
but since i'm doing bahasa indonesia into english i think its fine for me.
hope i can get to proper solutions but at least with the current state i'm pretty satisfied.
again this is very cool stuff.
I'm glad you found a solution!
You can indeed use the post processing option.
Please share with everyone your work so we can learn. 🙏
@@royshilk like i already stated, i pretty much copy paste based on image shown in your video, with a difference of :
{"text":"{{input}}", "source_lang":"ind_Latn","target_lang":"eng_Latn"}
Note:
For people that tried it the first time capital L on Latn are important the very first time i tried without considering case sensitive it doesn't work.
then what works for me to make test request working and the plugin to works are below:
Output Parsing Options:
Content-Type: Text
Regex:
Regex Flags:
Regex Group=0
Post process Regex: I'm not sure
Replace:
again i have 0 knowledge about how it works or what Regex even mean, but when i tried to copy according to the image on the video when i press test request it says in red text above the button as : Failed to match regex.
maybe i'm missing something or didnt read the documentation properly and i'm sorry for that, hopefully this comment somehow help with the development, cheers!
tldr i think i basically just tell the AI to replace its phrase "I'm not sure" which is the idle response if there is no additional input from localvocal(blank input) into [blank] if i understand it correctly
P.s: i tried english > indonesia too, and the idle response are berbagai(means several) and i did the same to it too(changing the AI response of blank input which is "berbagai" into [blank] or . . .
Hello, I download and installed it on windows, then open OBS and nothing came up. ??
there could be many issues... like your windows OS version or how you installed OBS. can you share a link to your OBS logs? you can get the link from the Help menu in OBS
@@royshilk No AJA devices found, skipping loading AJA UI plugin
22:20:02.380: Failed to initialize module 'aja-output-ui.dll'
22:20:02.425: No AJA devices found, skipping loading AJA plugin
22:20:02.425: Failed to initialize module 'aja.dll'
22:20:02.427: Skipping module '../../obs-plugins/64bit/chrome_elf.dll', not an OBS plugin
22:20:02.562: [CoreAudio encoder]: Adding CoreAudio AAC encoder
22:20:02.677: Skipping module '../../obs-plugins/64bit/ctranslate2.dll', not an OBS plugin
22:20:02.681: Failed to load 'en-US' text for module: 'decklink-captions.dll'
22:20:02.687: Failed to load 'en-US' text for module: 'decklink-output-ui.dll'
22:20:02.708: A DeckLink iterator could not be created. The DeckLink drivers may not be installed
22:20:02.709: Failed to initialize module 'decklink.dll'
22:20:02.875: Skipping module '../../obs-plugins/64bit/libcef.dll', not an OBS plugin
22:20:02.876: Skipping module '../../obs-plugins/64bit/libEGL.dll', not an OBS plugin
22:20:02.877: Skipping module '../../obs-plugins/64bit/libGLESv2.dll', not an OBS plugin
22:20:03.154: Skipping module '../../obs-plugins/64bit/libopenblas.dll', not an OBS plugin
22:20:03.286: [obs-browser]: Version 2.22.1
22:20:03.286: [obs-browser]: CEF Version 103.0.5060.134 (runtime), 103.0.0-5060-shared-textures.2594+gc69ad37+chromium-103.0.5060.134 (compiled)
22:20:03.692: [noise suppress]: NVIDIA denoiser disabled, redistributable not found or could not be loaded.
22:20:03.693: Failed to get NVVideoEffects.dll version info size
22:20:03.693: [NVIDIA VIDEO FX]: FX disabled, redistributable not found or could not be loaded.
22:20:03.774: LoadLibrary failed for '../../obs-plugins/64bit/obs-polyglot.dll': The specified module could not be found.
22:20:03.774: (126)
22:20:03.774: Module '../../obs-plugins/64bit/obs-polyglot.dll' not loaded
22:20:03.956: [obs-websocket] [obs_module_load] you can haz websockets (Version: 5.3.3 | RPC Version: 1)
22:20:03.956: [obs-websocket] [obs_module_load] Qt version (compile-time): 6.5.3 | Qt version (run-time): 6.5.3
22:20:03.956: [obs-websocket] [obs_module_load] Linked ASIO Version: 102800
22:20:03.989: [obs-websocket] [obs_module_load] Module loaded.
22:20:04.078: [vlc-video]: Couldn't find VLC installation, VLC video source disabled
22:20:04.177: ---------------------------------
LoadLibrary failed for '../../obs-plugins/64bit/obs-polyglot.dll': The specified module could not be found.
22:32:42.389: (126)
22:32:42.389: Module '../../obs-plugins/64bit/obs-polyglot.dll' not loaded
22:32:42.478: [obs-websocket] [obs_module_load] you can haz websockets (Version: 5.3.3 | RPC Version: 1)
22:32:42.478: [obs-websocket] [obs_module_load] Qt version (compile-time): 6.5.3 | Qt version (run-time): 6.5.3
22:32:42.478: [obs-websocket] [obs_module_load] Linked ASIO Version: 102800
22:32:42.496: [obs-websocket] [obs_module_load] Module loaded.
@@royshilk
Can you please actually copy in the http link to the logs from OBS? There's an option to get a link to the logs and if you send it I will get the full picture of what's happening
Hi, once download the models what do i do next, sorry for the silly question, i am so new at this and i want to translate for a church serviced in Vietnamese to English's. @@royshilk
where can i find the model which you are using in this video? i cannot find any suitable model for the API to work
compatible NLLB models are downloadable from forum.opennmt.net/t/nllb-200-with-ctranslate2/5090
Hello, I tried to set up the plug-in, the first time no problem, but I can't find the language code for French (I'm from France) so I cross my fingers and enter :
{"text":"{{input}}", "source_lang":"fr_Latn", "target_lang":"eng_Latn"}
(French is not in the readme list)
the translation is slow and does not include all words.
and the second time I launch OBS, I can't get it to work, even the "translation failed" error message doesn't appear.
But bravo for the work already accomplished
This is a list of all lang codes
huggingface.co/facebook/nllb-200-distilled-600M/blob/main/special_tokens_map.json
It's slow because it runs local on your computer
If you're missing words make sure the input includes all the words
Doesn't work, port 18080 gives an error.
Did you start the server? Is it (and not https)?
Hey, I was wondering where the language codes are for the body section? I checked the readme over a few times but cant seem to find them. Was also wondering if you can translate into other scripts besides latin? Thanks for the great plugin! :)
Here's a link where all language codes are described: github.com/facebookresearch/LASER/blob/main/nllb/README.md
It can definitely translate to much more than just Latin scripts...
Thanks 🙂@@royshilk
How can I use OpenAI chatgpt for licey translation with you plugin? Please explain where and what settings need to be made.
Check out obsproject.com/forum/resources/url-api-source-live-data-media-and-ai-on-obs-made-simple.1756/ that will help you connect chatgpt to OBS
@@royshilk Thank you so much
srry what is .spm i couldn't get
Sentence Piece Model. you can download from pretrained-nmt-models.s3.us-west-2.amazonaws.com/CTranslate2/nllb/flores200_sacrebleu_tokenizer_spm.model if you're using NLLB
YOU TELEPORTED ME TO 2134 THANKS @@royshilk
what input mean in request ? i can find any obs element have name like input and they make me hard to understand
I think I might be doing something wrong. My stream disconnects when I start the HTTP Server.
it work perfectly thank u
I'm glad it worked out for you! If you have a public creation you made with it let me know
@@royshilk now i try to make my voice show subtitles translated in real time
i did it but it's translate weirdly (its funny tho but i want do better)
I have the error after downloading and selecting "Failed to load CT2 model. Unable to open file 'model.bin' F:/*filepath*" I tried all three of the NLLB-200 multilingual model, but all of them gave me the same error. Any way you can help?
Did you point the settings to your models? Seems like the plugin simply isn't configured to find the model filed. Tomorrow I'm streaming "office hours" so you're welcome to join and ask your question and I'll help in real time!
@@royshilk ok sounds good! Thank you.
@@violet_turning_violent if you need help i'm live atm on my channel
Hello ! Thank you for your work. it seems that after uninstalling Polyglot on Windows, the language files stay on "AppData\Roaming\obs-studio\plugin_config\"
Right it's possible the uninstaller doesn't know about downloaded files and won't delete them.. not sure how to make it do that though. Would need to research
do you have a tutorial to install in windows? It looks very different on my end and I have no idea what I am doing. It isnt showing up things on my obs.
Thanks !
But it says for me "Failed to Load Model CT2 Model. File model.bin is incomplete: failed to read a buffer of size 262354944 at position 191"
I really don't understand what does that even mean-
Did the download complete properly? Sounds like the model download failed at some point
@@royshilk Yes it did, I redid it but it stayed the same-
discord.gg/DWKaM2JWNH im happy to look at logs
I'm getting Error
Failed to load SPM. Not found: model file path should no be empty
Did you download the spm and set up the file in the settings?
The SPM is downloaded from pretrained-nmt-models.s3.us-west-2.amazonaws.com/CTranslate2/nllb/flores200_sacrebleu_tokenizer_spm.model
When using this plugin, my CPU utilization reaches 100% (Windows 11 user). The system specifications include an i5-11400H processor, 16GB RAM, and an RTX 3050 4GB graphics card. taskbar showing: obs (23) and chrome (20).
Yes the local translation is heavy on the CPU. It's running an LLM on your PC which will be costly. on your machine I would recommend running a cloud translation. You can use OpenAI chatgpt for licey translation or a service like DeepL, which you can integrate with my URL/API Source
@@royshilkHow to Integrate Deep L
this might be a noob question, but i can't seem to find a Data source in obs (i'm using windows version, i see you're using Mac version).
Is your data source the same as Text (GDI+) on windows? or should I be using a Browser source instead?
Get URL API source obsproject.com/forum/resources/url-api-source-live-data-media-and-ai-on-obs-made-simple.1756/
Is there any program we can use locally to translate downloaded videos?
lexisynth can help do that obsproject.com/forum/resources/lexisynth-live-ai-for-speech-transcription-translation-and-synthesis.1902/
I don't see anywhere in the read me the language codes.
I get you're an engineer not a teacher, it's like a meme issue with engineers lol. But you should be more detailed with your walk throughs. Don't skip steps that are obvious to you. If you say something is somewhere, show it and how to get there.
Not trying to be rude, just I'm not even used to obs yet, and it's the 3rd video I've seen of yours where I'm just unable to follow you. And I'm pretty tech savvy ....
Thanks for the feedback! I'm trying to get better, but at the same time I value your time and I won't make videos with fluff and baits. The codes are indeed mentioned in the readme but here they are github.com/facebookresearch/LASER/blob/main/nllb/README.md
@@royshilk Thank you for adding those, because they weren't there before. But also, no non latin languages? I don't see Korean, Japanese, Or Chinese or anything of the sort. Thought it was no language left behind lol
@@davidmorton6406 The list given only contains less used languages added with LASER3, it doesn't contain the codes for english, french, italian, etc. either. You can find a more complete list by going to the side bar, opening the "laser_encoders" folder, and looking in the "language_list.py" file.
It's not clear to me how i am supposed to add "polyglot" to my "sources"
Polyglot is an internal service and not a source or a filter on its own
how can i find my http url ??
Hi, sometimes i get this error: error parsing json. How i can fix this?
you actually don't need JSON parsing - the output of polyglot server is clear text. so please just use "Text" for the parsing option (instead of "JSON")
@@royshilk yeah i use it but i have always this error messagge sometimes. I use cloud close captions and write captions in a text source. Then url source and setting all likes the video... If you want i can write to you on obs forum and send some screen if you need them
@@MannaggACardarellyes! I appreciate screenshot, also you can open an issue github.com/obs-ai/obs-polyglot/issues so I can track
The plugin is still waiting approval on the forum..
@@royshilk hello i have the same problem "error parsing json "
Please try the release github.com/occ-ai/obs-urlsource/actions/runs/6758096048#artifacts
Until I release the official v0.2.1
I'm just waiting on a final pull request valuation from another tester
Is there a model that supports japanese to english?
Yes the m2m100 model supports this pair. And you can probably find others more specific on hugging face..
@@royshilk How can I know what the abbreviation for Portuguese is: I would like to speak Portuguese and translate it into English. But I don't know what the abbreviations of the words would be.
here you can find all language codes github.com/facebookresearch/LASER/blob/main/nllb/README.md
@@royshilk Thanks I didn't pay attention to the model selection when I had first set it up. I ended up using MangaOCR to transcribe japanese characters to text in order to feed the model
@@Rubafix989 ive now added >10 new languages to the OBS OCR plugin (github.com/occ-ai/obs-ocr) so you could use that..
Is it best to close OBS before running your plugin install process, or does it matter?
Yes OBS needs to restart after installing the plugin. So it's best to close OBS and after installing to reopen OBS
what do i put for the model and spm ;-;
You need to download the model and SPM files first, and then you put in the file paths. The download links are in the description
hi how are you its very good plugin i am trying to install it but i do not know where i get the problem how i will take your help in this which communication source if you will be guide me i will be very thankful. after every thing its shows me the jason error
Hi. I'm able to help on discord.gg/H3YpY5nX . Here is inconvenient to help with JSON problems
I'm on version 0.2.5 url source and getting error parsing json when i'm using text. :(
You can reach out on discord/Patreon/X and I can do my best to assist. I need more information to see what's going on
my polygot doesn't work :::(
I'm sorry to hear that! Do you want to meet us on discord.gg/fzsGqdngAP for live support?
Does this output text that is "embedded" into the video stream, or is the user able to turn it on and off? And does it only translate to one language at a time?
you can output to a text source - in which case it would be "embedded", or you can select in URL/API Source to output the result to the current stream as captions - in which case your viewers can turn on / off
right now it's single language, but if you use something like OpenAI LLM API you can do multiple language at once, just ask the GPT that
thanks!@@royshilk
how to get code for engish to creole
Here you will find all language codes github.com/facebookresearch/LASER/blob/main/nllb/README.md
How accurate is it?
Depends on the model you use and what you are looking to translate. Simpler sentences are translated very well
I downloaded the model from the link and there doesn't seem to be an SPM file.
Nvm.. found it, but it still isn't working.
WHat?? No french language code?
Definitely there's French.. it's a major language. Do you want to reach out on our support discord server discord.gg/kp4QgSzt ?
Neat, but not very practical.
Thanks for the feedback. How can I make it better and more useful? What would you want to see?
@@royshilk I have a dutch friend who wants to show real time english subtitled translation for english viewers. Maybe this is already possible? but i'm struggling to achieve that at the moment. Using your guide but not seeming to get it working.
I'm happy to help get the translation up and running
You can find me on the OBS discord server, or DM on X/Patreon etc.
Also I'm live streaming and helping people with plugin problems twice a week TuTh, you can join that
@@royshilk That's great. Thank you.
looks to complex
not for the faint of heart :DD how can i make it simpler?