Tool RPGM DazedMTLTool - A tool that provides quality MTL translations using ChatGPT

[Cutefunniness]

Does it support 18+ ?

If you mean uncensored/unrestricted lines then it yes depending on what provider the api uses (G4F has many providers aka sites that have chatgpt and other models, some unrestricted and such).

I've already replied to another user explaining how to customize which provider to use when using G4F to get the best results, but do note that G4F may fall back to the providers that give the "I can't do that" message if the providers providing the unrestricted/uncensored api are timing out due to either request limit or unavailability.

Do note that this is a temporary fix and some of the providers I disabled are either not working or erroring out, the next update to G4F will potentially fix and improve many of the providers hopefully!

Give it a try, you'll have to retry many times when translating a file (unless you're super lucky and get it all done once haha) until it's all done as DazedMTL noted in their description. Yup!

 

[revyfan]

Neat things dropped.

First, google dropped another "experimental" gemini model, just called '1114'. It's a weird one, because it seems surprisingly competent and seems more likely to write erotic material, but for translations it's only slightly better than my experience with pro 2... One thing that actually gets me, though, is it's uncanny ability with vision. Not only can it get more Japanese characters right compared to pro 2, it almost seem like it can type out manga text in the right order (Something pro 2 could not do) around 80% of the time. Honestly, if you're someone who's been wanting to translate Manga/doujins/CGs and don't quite have the skills to properly type out the Japanese text, then it is one of the best 'free' (With limits) model to use right now. The only real downside is google going from models with a million context, to a model with only 32k. Some speculate it's because this was Gemini Ultra, but I don't really know.

Now, onto a big release! Mistral Large 2.1 AND, my absolute new favorite model, Pixtral large. Last time I ever used a Mistral model, it didn't really impress me, but PIXTRAL LARGE is just fucking perfect for erotic translations. Seems like it can do a good job with most context, sometimes it does kind of mess up, mostly with slang, but if you're even slightly paying attention to the context, it's pretty easy to steer it back!

 

[Viir]

Hey @Cutefunniness. I've been using the GPT4F that you have posted about and have managed to get it to work. but for some reason its frequently spits out some automated message in Spanish whilst deleting the text I want to translate. Is there something i can do to fix this?

 

[jaden_yuki]

did anyone test the new aya expanse? everyone disables the huggingface serverless inference, and the cohere's free rate limits suck

 

[Beyo]

What other engines outside RPGM does this work with? I'm curious to know if it's compatible with Nscripter now.

 

[Cutefunniness]

Hey @Cutefunniness. I've been using the GPT4F that you have posted about and have managed to get it to work. but for some reason its frequently spits out some automated message in Spanish whilst deleting the text I want to translate. Is there something i can do to fix this?

Glad to hear you got it to work and yes there's away to get rid of the automated spanish messages, you'll have to edit out the "ChatGptEs" provider by following my post here:
https://f95zone.to/threads/dazedmtl...anslations-using-chatgpt.155635/post-15329625

You can customize the provider list as well to improve getting the more spicy/explicit lines translated!
(It'll take a while but the chance of your lines getting translated increases so you'll eventually get it done!).

Oh and be wary that G4F has updated so the list I made before in my post may have providers that aren't available anymore...actually I just updated my post to include that information incase someone who's not aware uses that list!

Hope this helps!

 

[Viir]

Glad to hear you got it to work and yes there's away to get rid of the automated spanish messages, you'll have to edit out the "ChatGptEs" provider by following my post here:
https://f95zone.to/threads/dazedmtl...anslations-using-chatgpt.155635/post-15329625

You can customize the provider list as well to improve getting the more spicy/explicit lines translated!
(It'll take a while but the chance of your lines getting translated increases so you'll eventually get it done!).

Oh and be wary that G4F has updated so the list I made before in my post may have providers that aren't available anymore...actually I just updated my post to include that information incase someone who's not aware uses that list!

Hope this helps!

Hey thanks for the reply and the help. I will probably message you again because I'm still fairly new to using API and python programing/terminal.

Anyway you posting about G4F was helpful for me as the translation projects I'm pursuing would have cost way more than i thought it would ( I hoped it would be somewhere between £5-£20 and instead I'm getting £35-£60 which is a bit of a surprise).

So thank you for bringing this up.

Re-Edit: Hey managed to fix the issues that cropped up and there is really only one problem left. In a fair amount of instances (55%) the tool chooses to give me a context or analysis on the line to be translated instead of a translation itself is there anyway to fix this? Cutefunniness

 

[Cutefunniness]

Hey thanks for the reply and the help. I will probably message you again because I'm still fairly new to using API and python programing/terminal.

Anyway you posting about G4F was helpful for me as the translation projects I'm pursuing would have cost way more than i thought it would ( I hoped it would be somewhere between £5-£20 and instead I'm getting £35-£60 which is a bit of a surprise).

So thank you for bringing this up.

Re-Edit: Hey managed to fix the issues that cropped up and there is really only one problem left. In a fair amount of instances (55%) the tool chooses to give me a context or analysis on the line to be translated instead of a translation itself is there anyway to fix this? Cutefunniness

Glad to help and wish you the best on your translations!
As for the issue of context/analysis, that's because of Blackbox and apparently because of it being overwhelmed with responses, it's not working right and thus system prompts aren't going through and such.

I reported this on G4F github and that's the answer I got as I was suffering the same issue!
You'll just have to deal with it...yup!

 

[Viir]

Glad to help and wish you the best on your translations!
As for the issue of context/analysis, that's because of Blackbox and apparently because of it being overwhelmed with responses, it's not working right and thus system prompts aren't going through and such.

I reported this on G4F github and that's the answer I got as I was suffering the same issue!
You'll just have to deal with it...yup!

Cutefunniness
Hey it's me again. Just wanted to see if you could answer some questions I had regarding coding and errors?

1.) I have a hunch that G4F is not able to read the prompt.txt file from the Dazed repository and that when the text we want translated is going through to the providers the AI simply assumes what we want (e.g Analysis, context, etc almost like a slot machine in a way). I could also be completely wrong (I probably am). I'm just trying to think of the reasons for why some things are translated whilst others are not.

2.)Now I've edited extractTranslation like you explained (correctly hopefully...) but one thing I noticed is that on the dazed tool side i get the following error (extractTranslation Error: No valid JSON content found on String 申し訳ありませんが、そのリクエストにはお応えできません). Now i have translated that Japanese (Sorry, we cannot accommodate that request) but I don't get why it's in Japanese when Dazed and it's translations are set for English. Anyway just wondering if there's a fix for this and what JSON content is valid for extractTranslation?

3.) Is there any way to fix the issue with the Proactor event loop error that pops up?

Anyway thank you once again for responding to me on this topic and answering my questions. As with me being quite the greenhorn when it comes to programing i can see it being fairly annoying after a while so thank you for being patient with me.

All in all in the months that I have been trying to learn how to program and translate i have really come to appreciate the work that people like you, dazedanon and the G4F devs have done.

Take care now.

 

[Cutefunniness]

Cutefunniness
Hey it's me again. Just wanted to see if you could answer some questions I had regarding coding and errors?

1.) I have a hunch that G4F is not able to read the prompt.txt file from the Dazed repository and that when the text we want translated is going through to the providers the AI simply assumes what we want (e.g Analysis, context, etc almost like a slot machine in a way). I could also be completely wrong (I probably am). I'm just trying to think of the reasons for why some things are translated whilst others are not.

2.)Now I've edited extractTranslation like you explained (correctly hopefully...) but one thing I noticed is that on the dazed tool side i get the following error (extractTranslation Error: No valid JSON content found on String 申し訳ありませんが、そのリクエストにはお応えできません). Now i have translated that Japanese (Sorry, we cannot accommodate that request) but I don't get why it's in Japanese when Dazed and it's translations are set for English. Anyway just wondering if there's a fix for this and what JSON content is valid for extractTranslation?

3.) Is there any way to fix the issue with the Proactor event loop error that pops up?

Anyway thank you once again for responding to me on this topic and answering my questions. As with me being quite the greenhorn when it comes to programing i can see it being fairly annoying after a while so thank you for being patient with me.

All in all in the months that I have been trying to learn how to program and translate i have really come to appreciate the work that people like you, dazedanon and the G4F devs have done.

Take care now.

Let me help!

1. About G4F and prompts, I learned that some of the providers can't use the system prompt but some can thus some are able to read the prompt. I haven't tested them all but I'm aware that some providers aren't capable of reading system prompts as they don't support it (like ChatGpTEs for example).

I'm currently dealing with an issue with versions of G4F past v0.3.6.6 and above as they're not working as intended for me. One of the errors, 422 Unprocessed Entity, has been fixed but an unknown error is causing G4F to not work with DazedMTL, causing 500 Internal Server errors despite providers like Blackbox, chatgpt, etc being avaiable.

Have you upgraded and had any of these issues?
I'm currently using v0.3.6.5 of G4F as it's the last version that's working with DazedMTL.
Currently getting help on figuring out the issue on their github.

2. That line that's in Japanese is from the Blackbox provider, it does that because it's overwelmed and not working as I described before. This isn't on G4F but on the provider's side themselves, you can choose to disable the provider or ignore it.


3. That's something G4F does randomly and one I don't know anything about but...you can just ignore it, it hasn't caused me any issues despite the error appearing. Yup!

Hope this helps! ^^

 

[Viir]

Let me help!

1. About G4F and prompts, I learned that some of the providers can't use the system prompt but some can thus some are able to read the prompt. I haven't tested them all but I'm aware that some providers aren't capable of reading system prompts as they don't support it (like ChatGpTEs for example).

I'm currently dealing with an issue with versions of G4F past v0.3.6.6 and above as they're not working as intended for me. One of the errors, 422 Unprocessed Entity, has been fixed but an unknown error is causing G4F to not work with DazedMTL, causing 500 Internal Server errors despite providers like Blackbox, chatgpt, etc being avaiable.

Have you upgraded and had any of these issues?
I'm currently using v0.3.6.5 of G4F as it's the last version that's working with DazedMTL.
Currently getting help on figuring out the issue on their github.

2. That line that's in Japanese is from the Blackbox provider, it does that because it's overwelmed and not working as I described before. This isn't on G4F but on the provider's side themselves, you can choose to disable the provider or ignore it.


3. That's something G4F does randomly and one I don't know anything about but...you can just ignore it, it hasn't caused me any issues despite the error appearing. Yup!

Hope this helps! ^^

Thanks again for the response. By the way to answer on whether i was suffering from the (version, 422, and 500, errors) then the answer is yes but this was after i tested and used version 0.3.6.5 so when i updated past that to 0.3.6.7 and it stopped working i downgraded back down.

So it is not just you and if you want me to test things on my end let me know.

 

[dazedanon]

Guide has been updated to closer match my current process as well as fix the images.

 

[Cutefunniness]

Thanks again for the response. By the way to answer on whether i was suffering from the (version, 422, and 500, errors) then the answer is yes but this was after i tested and used version 0.3.6.5 so when i updated past that to 0.3.6.7 and it stopped working i downgraded back down.

So it is not just you and if you want me to test things on my end let me know.

Ello, I have an update with G4F! After some trial and error figuring out the issue, I was able to figure out what was causing it and reported it on github yesterday and now the newest version of G4F fixes everything!

Starting with v0.3.8.0, many new fixes and provider updates have been added along with fixing the main 500 internal server error issue! Also new updates to ChatgptEs and Blackbox have been made to fix/add system prompts so you can try using it again! I'm testing it out right now and my gosh things are a whole lot better!

They're new providers as well so when you decide to modify the models.py list, I recommend testing out the newly updated/fixed providers first before deciding on what to disable!

Also I like to add that there's a provider that the devs at G4F recommend trying out that was just added to the newest verison, Microsoft's Copilot which is apparently built off of GPT-4o so you can try that out and see how it works!

Copilot can be used with "gpt-4" so switch your .env's model to that along with modifying the module's py file to target gpt-4 as well if you decide to use it! (You can add another entry that targets either gpt-4 or gpt-4o by copying the elif, that way you can just modify the .env file whenever you decide to switch between one or the other).


OH and a o1-preview and o1-mini provider has been added as well, I haven't tested it out nor know if they support system prompts but you can test that out too!

Happy thanksgiving/happy holidays to you and dazedanon and enjoy translating away!

 

[jaden_yuki]

Ello, I have an update with G4F! After some trial and error figuring out the issue, I was able to figure out what was causing it and reported it on github yesterday and now the newest version of G4F fixes everything!

Starting with v0.3.8.0, many new fixes and provider updates have been added along with fixing the main 500 internal server error issue! Also new updates to ChatgptEs and Blackbox have been made to fix/add system prompts so you can try using it again! I'm testing it out right now and my gosh things are a whole lot better!

They're new providers as well so when you decide to modify the models.py list, I recommend testing out the newly updated/fixed providers first before deciding on what to disable!

Also I like to add that there's a provider that the devs at G4F recommend trying out that was just added to the newest verison, Microsoft's Copilot which is apparently built off of GPT-4o so you can try that out and see how it works!

Copilot can be used with "gpt-4" so switch your .env's model to that along with modifying the module's py file to target gpt-4 as well if you decide to use it! (You can add another entry that targets either gpt-4 or gpt-4o by copying the elif, that way you can just modify the .env file whenever you decide to switch between one or the other).


OH and a o1-preview and o1-mini provider has been added as well, I haven't tested it out nor know if they support system prompts but you can test that out too!

Happy thanksgiving/happy holidays to you and dazedanon and enjoy translating away!

I'm curious about the o1 provider

 

[Viir]

Ello, I have an update with G4F! After some trial and error figuring out the issue, I was able to figure out what was causing it and reported it on github yesterday and now the newest version of G4F fixes everything!

Starting with v0.3.8.0, many new fixes and provider updates have been added along with fixing the main 500 internal server error issue! Also new updates to ChatgptEs and Blackbox have been made to fix/add system prompts so you can try using it again! I'm testing it out right now and my gosh things are a whole lot better!

They're new providers as well so when you decide to modify the models.py list, I recommend testing out the newly updated/fixed providers first before deciding on what to disable!

Also I like to add that there's a provider that the devs at G4F recommend trying out that was just added to the newest verison, Microsoft's Copilot which is apparently built off of GPT-4o so you can try that out and see how it works!

Copilot can be used with "gpt-4" so switch your .env's model to that along with modifying the module's py file to target gpt-4 as well if you decide to use it! (You can add another entry that targets either gpt-4 or gpt-4o by copying the elif, that way you can just modify the .env file whenever you decide to switch between one or the other).


OH and a o1-preview and o1-mini provider has been added as well, I haven't tested it out nor know if they support system prompts but you can test that out too!

Happy thanksgiving/happy holidays to you and dazedanon and enjoy translating away!

Now thats some bloody good news and I can’t deny that this has put me in a merry mood!

Thanks again for all your help.

 

[jaden_yuki]

Ello, I have an update with G4F! After some trial and error figuring out the issue, I was able to figure out what was causing it and reported it on github yesterday and now the newest version of G4F fixes everything!

Starting with v0.3.8.0, many new fixes and provider updates have been added along with fixing the main 500 internal server error issue! Also new updates to ChatgptEs and Blackbox have been made to fix/add system prompts so you can try using it again! I'm testing it out right now and my gosh things are a whole lot better!

They're new providers as well so when you decide to modify the models.py list, I recommend testing out the newly updated/fixed providers first before deciding on what to disable!

Also I like to add that there's a provider that the devs at G4F recommend trying out that was just added to the newest verison, Microsoft's Copilot which is apparently built off of GPT-4o so you can try that out and see how it works!

Copilot can be used with "gpt-4" so switch your .env's model to that along with modifying the module's py file to target gpt-4 as well if you decide to use it! (You can add another entry that targets either gpt-4 or gpt-4o by copying the elif, that way you can just modify the .env file whenever you decide to switch between one or the other).


OH and a o1-preview and o1-mini provider has been added as well, I haven't tested it out nor know if they support system prompts but you can test that out too!

Happy thanksgiving/happy holidays to you and dazedanon and enjoy translating away!

I was testing g4f with the web GUI and I did find it quite slow. Is it like this normally? maybe the stream endpoint is slower? or my potato PC perhaps

 

[mmoo30]

I followed your tutorial step by step but got this error when trying to run the "start.py":
[Input: 0][Output: 0][Cost: $0.0000][20.1s] ✗ Client.__init__() got an unexpected keyword argument 'proxies'

after this, the files in files folder go to the translated folder, but nothing has changed inside them.

Do you have any idea whats the issue?

 

[DariusPentakill]

How big is the difference in quality between using gpt 4o and gpt 4o mini ?

 

[dazedanon]

How big is the difference in quality between using gpt 4o and gpt 4o mini ?

Pretty big 4omini makes a lot more mistakes especially with genders and subjects.

 

[DariusPentakill]

Makes sense, considering the price difference. Makes it hard to use 4o for entire games. But the results i've gotten from 4o mini are still miles ahead of standard mtl. How close is 4o to a manual translation do u think?