itch.io

Hi pal!

Sorry, this is a bug that has completely passed me by because I always had the model downloaded. I will make sure that this is fixed for the next alpha version.

Thank you very much for making this post. I'll let you know when the new version drops, should be tonight or tomorrow. :)

╰(*°▽°*)╯

Very much appreciated by you, good to see people using the app helping each other out! :)

Thank you! I fixed a bug in the latest version that didn't let other users download updates to the presets they had already downloaded. Must have broken it when I structured everything up a bit after deciding which style I was going for.

Ah, no worries! There's definitely value in leaving them up, just try to make sure they're not tagged as a free and it's perfect! :)

That's interesting. I haven't heard or seen this CLI tool. How fast was the Flash model through that method?
Yeah it sounds like a bit of a setup, probably something the more advanced users could appreciate though. I wish I had more time to try out all sorts of APIs and LLM models.. lots of interesting stuff out there.

Thanks! Was a headscratcher for me when I saw this haha.

Shoot, sorry, my bad, I must have done the testing on the Preview version like Norby suggested.. I have fixed it now, I should be uploading a new version tonight. :)

No worries, sorry that I created an unnecessary bug hah.

By the way - I fixed the $.. being formatted to $. in the JSONPath in 0.6.1-alpha.1, so I'd recommend updating your Azure presets now to use $.. instead of $. since the latter is an invalid path for the translation results for Azure. :)

And oh, I finally managed to create an Azure account. Seems pretty good!

Hi pal,

The quotation mark bug should be fixed in the new 0.6.1-alpha.1 version. I also reverted how architecture is identified for the release build (the change I made for development testing), so Attached mode will now work just like before. If it still doesn't work for you, let me know, because then it was unrelated to that change.

Hi,

Thank you for reporting these issues.

I had not noticed the issue with quotation marks. I thought that bug was squashed months ago, but it is possible I reintroduced it while making changes to related code.

Does Attached not work at all for any game? I made some changes earlier to make testing easier for myself, that is likely the cause of it. However, I haven't had any issues with Attached mode myself since I made the change. I'll investigate it further tomorrow.

I have also noticed that the 'fix' to always make the JSONPath start with $. (only one dot) is breaking the path for Azure. I didn't realize some APIs are setup like $..

I'll get that fixed too. :)

That's great to hear, thank you for letting me know!

I'll likely go through some keyboard layouts at some point in the future to get these all kinds of non-supported keys manually added this way. Since I am unable to test these out myself at the moment, I'll leave it as a on-request thing to add, otherwise we may end up with a bug related to how these keys are read.

0.6.0 is now re-released with the fix :)

Oh my god... I put {} as the placeholder identifier instead of %%.

If you put {api_key}, {text} etcetera it should work. I'm reverting it back to %api_key% now, will be uploading 0.6.0 again shortly.

🫩🫩

Thank you! If I remember correctly now, before 0.12, I had missed to commit some changes in a shared library, accidentally discarded it and then had to rewrite it and I must have broken it. I did this at some point anyways, but sounds like it must have been before 0.12. I'll fix this and reupload 0.6.0 everywhere in a few hours.

I'm not home yet for another few hours. Is this only happening with Azure?

Crap. I don't remember making any changes to this apart from fixing an API key bug when using the quick apply dropdown..

I'll have a look today, sorry for the inconvenience.

Thank you very much! I'll get through testing them a little bit today. (☞ﾟヮﾟ)☞

I think that prompt will do, if anyone reports any issues with it in the future, it's always easy to edit again :)

Interesting - maybe it's just something with my connection then, but it's only bad with the non-flash and non e2b and e4b models for me.

I'd personally say that seconds slower is a vast difference in experience 😅 In my opinion anything slower than the internal models is barely passable for me, but I'm well aware a lot if not most people care more about 100% accurate translations than they do about speed.

Sounds great! :)

Oh, don't worry about that. I don't expect things to be straightforward all the time, quite the opposite, hence why feedback is super important.

Did you try it with the gemma-3n-e2b-it model and gemma-3n-e4b-it models?

I think the other Gemma ones might work better with the occasional miss, but they are extremely slow, sometimes a few seconds, sometimes 40 seconds. Only the flash ones seem usable of the non-pro models, unfortunately.

I would still strongly suggest to change it, putting the to-be-translated text in the middle of the prompt will result in errors. :)

Ah! I didn't check the Azure ones. I just saw that a good few of the new presets didn't utilize it when users want to change the models.

Also, you probably realize this when you double check, but you can just use %target% for the Azure API endpoint instead of adding a redundant %language% placeholder. Otherwise I think things seem to look fine, I haven't had time to test the new ones though. I'll try to do so today.

*Edit, sorry, I thought I had changed it to work for endpoint and headers, but I never did... I'll fix that for the next version.

Did you test it with that exact paragraph? It usually only starts breaking when the text you try to translate has any resemblance to an actual prompt message. In this case it is talking about the English language, which the LLM reacts to as if it is a continuation of the prompt rather than a text to translate

Yeah, no worries, I'm very grateful that you've put effort into this, it is something I should have setup myself a long time ago but I'm low on time myself and also intrigued to see how a community repository can evolve on its own. :)

Thank you for uploading all those new presets!! ⭐

I saw that you didn't utilize the custom placeholders. Did you not get it to work or did you not see that part in the creation window?

Sweet!

Yeah, it may seem like it is working perfectly, but there's a problematic part about the way you structured the prompt. The AI will have a hard time understanding it if the text to be translated isn't separated in some way from the prompt text itself.

For example, the text below won't translate with the prompt you put;

Request
English texts for beginners to practice reading and comprehension online and for free. Practicing your comprehension of written English will both improve your vocabulary and understanding of grammar and word order. The texts below are designed to help you develop while giving you an instant evaluation of your progress.

Result
Okay, I understand. Please provide the English texts. I will translate them into natural-sounding Swedish, focusing on accuracy and fluency, and providing only the translation.

This prompt below works just fine, although it could definitely be structured better too;

"Translate the following %source% text to %target%. Pay attention to accuracy and fluency. You are only to handle translation tasks. Provide only the translation of the text. Do not add any annotations. Do not provide explanations. Do not offer interpretations. Correct any OCR mistakes. Text:\n\n%text%"

Request
English texts for beginners to practice reading and comprehension online and for free. Practicing your comprehension of written English will both improve your vocabulary and understanding of grammar and word order. The texts below are designed to help you develop while giving you an instant evaluation of your progress.

Result
Engelska texter för nybörjare att öva läsning och läsförståelse online och gratis. Att öva din läsförståelse i engelska kommer både att förbättra ditt ordförråd och din förståelse för grammatik och meningsbyggnad. Texterna nedan är utformade för att hjälpa dig att utvecklas samtidigt som du får en omedelbar utvärdering av dina framsteg.

Structure is key for LLMs, one tiny mistake could be devastating for the results! :)

Hey bud,

I believe I fixed so that you'll be able to see the actual error now when you try to upload. Please let me know if it is still stopping you.

Don't worry about it, but I temporarily downvoted the non-flash Gemini presets due to them not at all generating the translated text properly when testing with paragraphs. Not sure if the models are flawed or if the prompt is. Also, I can personally not make any requests to the Pro versions, as it says that I hit the limit, so I presume they have no free requests at all.

Thank you for putting in effort into this. I'm going to upload some more presets to the LLAMA repository tomorrow, might hit a few Custom API ones too. :)

Haha, well, you and me both need to nail down our sleeping patterns then.

Great!

Oh. That's definitely a bug. My best guess is that it is an unknown error that just falls under a blanket error that thinks you're offline. I'll take a look at it in the morning! :)

Ah okay, got it!

I went ahead and implemented a way to add custom placeholders in your presets, which will mostly solve the differentiating model issue, as well as regions etcetera.

I'm refraining from adding a new section for api key/models/limits etcetera at the moment. I feel like it should be possible to get most of that information nicely input into the description now that users can click on links & select text from it, and directly paste that into the placeholder input.

I did just get an idea though - it would be cool for the user who creates the preset to be able to 'tag' the model name in the description, which then could be clicked on to automatically be added to a custom %model% placeholder that the user created. Or maybe even making it so that if you tag a line of text in the description, it will show up as a selectable dropdown option connected to the placeholder input.. Not sure if I'm making myself understandable here lol, I'm sleep deprived.

Cheers for the feedback, I'll take a look at making that text more legible. (I really need to get a higher resolution screen..)

Thank you very much for adding these. Please let me know if you find any bugs with the placeholders, but it should work fine afaik.

Ah! I'll make sure to increase the character limit. Might increase the upload limit too, it's just there to avoid abuse and ensure users' safety as much as possible.

Hmm. I do agree that each section ought to have its own description and URLs.. Just trying to think of a good way to implement this so it doesn't look more overwhelming than actually just having one single description and allowing the preset creator to just create links directly in that one description. Otherwise we'll end up with a lot of input boxes on the same page.
Could you elaborate on the models? Not sure I fully understood what you meant. :-(

No worries about the %% placeholders, I noticed, I just forgot to let you know haha. Thanks!

Ahh, yes! Have you found any other edge cases of placeholders being needed? I'm thinking that there may be many, in which case I believe I should add optional generic placeholders like %placeholder_1%, %placeholder_2% or similar to alleviate the issue instead of adding edge case upon edge case.

Thank you for the suggestions, very helpful! :)

Ahh, it was likely because I forgot to confirm the production deployment to the backend yesterday after I released the new update. Only did it a good 4-5 hours after. Or were you still on alpha.7 when you couldn't edit them?

Hey bud, I'll respond to this as soon as I wake up in the morning :)

Just one question - did you personally delete all the presets you uploaded?

Saw that you added loads, nice!! :)

I'm fixing a bug with the text output path at the moment, as well as adding an input for API registration URL + making it easier for users to select languages and api key etc when clicking apply. Please let me know if there is anything in the process you'd like to change!

Thank you very much,

Hey norby,

the 0.6.0-alpha.7 version has a community repository for Custom APIs. If you'd like to upload any of these presets onto it, that'd be cool! If you don't have the time or will, I'll get around to doing it very soon.
Thank you once again for making this gigantic thread, it's been most helpful to a lot of users.

Hi mate,

The latest 0.6.0-alpha.7 version might work for the NonConvert key. You should be able to at least set it, I am not able to test it since I don't have that key. Please let me know if it works! :)

Thank you.

Hi bud and thank you! :)

I recently had to switch to using physical keys instead of virtual keys due to updating to a newer version of IMGUI, which is what caused this key to be unsupported. I will look into the possibility of adding it for 0.6.0. I'll keep you updated. Thank you for filing the bug report!

Hi bud,

I do not have a specific ETA at the moment. I'm working on a pretty large update to streamline the ease-of-use of the app and really want to get it done and tested before it moves to Steam.

Do you have trouble buying it on Itch.io, or do you just want to buy it on Steam? :)

I'm sorry about the slow answer.

No worries pal!

Yeah, this sounds like a very peculiar issue. There must be something that some Unreal games have in common that is interfering. To further troubleshoot, I do need to know the names of the affected titles. Any Unreal game I have tried works fine on my device. Feel free to put them here or join the GameTranslate Discord and write to me there! :)

Hi pal,

This is planned. No ETA on when it would be added though. Haven't done enough research to know how much work it involves to implement it.

Niceee! ╰(*°▽°*)╯

I believe the issue with setting the correct keybinding should be fixed in 0.5.9. It isn't mentioned in the changelogs because it is a bug that came with the changes made in the alpha versions. :)

Thank you for providing context about the Pause/One Shot. Most helpful as always. Appreciate ya! ( ( •_•)>⌐■-■

I can confirm that I can reproduce the keybinding issue. Thank you very much for that, I'll have to wire it differently from the frontend.

The second issue could very well have something to do with the modifications I made to run Stream Friendly completely async (It was done to ensure minimum 60fps rendering at all times). I'll have a look at it eventually. Is there any particular reason you pause/unpause instead of using the 'One shot' key?
Thank you! :)

The tutorial only works with Spanish because I haven't had any time to localize the tutorial content. You cannot download the model because it is already included with the installation, and I do not provide a delete button because that would break the tutorial. :)

The default keybinding for closing windows is 'Escape'. If you hover a specific window, it will close that individual window, otherwise it will close all windows.

Let me know if you have any other questions/feedback, thank you!

Yep, the image recognition part is definitely the trickiest to perfect. I'm actually working on a complete test-bed behind the scenes, which should over time improve the OCR across the board, and potentially push forward a fullscreen solution! :)

Yep, that's on me, haha. I had an unintuitive and bad setup. I hope the recent update made it easier for you to test things out. The next update will make it possible to change these values from the dashboard in the tool too, and more improvements like that are on its way.