I had briefly looked at it. I don’t see substantial improvement with translation task, and it’s yet exorbitantly expensive i.e. with around $80 per million tokens, it would cost at least 10 times more than their 4o model, and is still in early preview. May be once they release a flash/mini model of this, I will corporate it.
Thanks for your feedback and I’m glad that you are finding this of help.
Regarding your issue, it appears to be a file encoding issue. What you say that “reedit the file again”, what editor you are using to edit the file. Is it that you are upload the file again to my website and editing it or using editor of your choice? If it’s related to my website, I can definitely look into why the encoding is being removed/changed.
Sorry for the trouble and late response. Are you still having this issue – as I can’t reproduce this issue. Did you get any error when this happen? If you can share a bit more detail about the flow/error, I can see what went wrong there and how to best fix it.
Hi. I often translate from English to Turkish. But I have an issue. The meaning of the “okay” word is “tamam”. So, “Okay my love”, “Tamam aşkım”. Well, your translation says “Tamam aşkım” when I upload the subtitle the definition is “okay” in context. Actually, must be only “Tamam.”
Sorry for the trouble Murat. From your description, I think the problem is with underlying translation service. Just out of curiosity, what translator you using. Did you try to use any other one and see if that helps.
Could you add a regular text/document translator as well? I often find ChatGPT just skipping over parts – with the paid subscription as well. I suppose if you did work through the API, it’d behave normally. And you’d also get the option to use other LLM models. So basically the same idea as the sub translator, just with attachments and/or a huge text field. Pricing could work the same. I really love this translator by the way, its working so good. The prices are also super affordable. Really appreciate the love and effort you put into this project, it shows and its amazing.
Hi Peter, thank you for feedback, and nice comments. I’m glad that you are finding this tool mostly of help.
You have a really interesting idea to support just normal text translation. No one has asked this before, mainly because most of these newer LLMs now support this feature natively. Though I can see your concern with skipping part of translation – I have seen this myself during few API calls. I will think about tweaking the interface to support just text based files. Though it’s not as simple as it first sounds, so I may not be able to do this pretty soon.
When translating from Hebrew to English, what’s the best way to avoid the punctuation (periods, commas, question marks, exclamation points, etc.) being placed at the beginning of the line, like it is in the original Hebrew, instead of at the end like it should be in English?
After translation, change the language alignment to “Right” (above the Translated Text table column). Also, please make sure “Embed Language Bi-Directional codes” is turned ON under “Save Settings”, so that it plays fine your preferred media player.
That doesn’t seem to do the trick. At least using GPT-4o, it ends up with most of the lines having the punctuation in the right place, at the end of the line. However, some lines still have it at the beginning.
Thanks for the link. Interesting! I think the problem is with the subtitle itself, where it has applied language direction too. Like if you apply Right-to-Left style on the content, you will notice how punctuation parts are reversed, and that’s probably why all these popular translators are failing, because they don’t expect Hebrew text with direction applied.
Actually, I did just try it with the Gemini Pro free, and it did seem to work okay at least for the first 50 lines. DeepL doesn’t have Hebrew as an option. I suppose I can try to do the full translation with Gemini, but I do prefer GPT-4o.
I would also suggest to try Gemini Flash 2.0 which seems to be doing far better results with punctuation and is slightly cheaper too. I’m still working on making it work with GPT-4o by tweaking the prompt to handle this properly.
I don’t know Hebrew, so can’t be of much help in determining how much effort it’s to fix this manually. If it’s just the punctuation symbols, moving these around would not be that tricky, but I’m afraid it might be more tricky/complex than just this.
Oh, I see. I just thought when you said “tweaking the prompt to handle this properly” that I could possibly put something in the prompt extension to handle it.
Yeah, I was taking about tweaking the prompt I internally use for the translation batches. Though that said, you as user can also tweak the prompt instruction by setting “Additional instructions” field, but this doesn’t seem to make much difference in this particular translation tweak.
Thanks James! This is already in my queue. However, they are not currently giving new API keys to the developers. As soon as their services are back up, I will integrate that.
Hello Akbar! I am currenty unable to log in with my google account, tried on different browsers and OS, it always returns with the same error message:
You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax…
Hi
Hiw can i download subtitles in extension.en.srt or .id.srt in save all
Because i need in this and subtitles translator downloads in full names like English Indonesian when downloading as save all in zip
When downloading multiple files, the ZIP file is the only option. I will think about adding some UI to let you individually download the files as well.
I don’t think currently any machine based translator would give you 100% accuracy for any type of content. I have found that Gemini and OpenAI based translators gives pretty good translation with maintaining context (around 80-90% accuracy), but none of these are prefect and in most of the cases you would need minor editing and tweaks to the translated subtitles to make final translation reach a satisfactory or human translation level.
Hello. Many thanks for this brilliant tool. Is there a way to set the source language. I understand the LLM figures it out 90% of the time. However, there are times, particularly if lines are short, when it mistakes the source language and therefore gives a different meaning in the target language.
True, it definitely helps when the source language is provided. For the LLM translators (Gemini, OpenAI) you can select the source language after you click the “Calculate Cost” button. It’s the first option in the configuration with title “From Language”. I hope this helps.
Sorry Patrick for the trouble. OpenAI models sometime skip the translation for few of line blocks and I’m already working to fix the properly. In the meantime, once the translation is done, you can re-translate a batch of lines to fix any partial omission or misalignment. You can more about it here: https://blog.syedgakbar.com/subtitles-translator-help-tips/#retranslate-block
PS. I have also initiated a full refund of this bad translation and you should have the money in your translation balance right away.
Have you tried GPT 4.5 yet? Is it worth adding or not much better than something like 4o?
I had briefly looked at it. I don’t see substantial improvement with translation task, and it’s yet exorbitantly expensive i.e. with around $80 per million tokens, it would cost at least 10 times more than their 4o model, and is still in early preview. May be once they release a flash/mini model of this, I will corporate it.
Regards, Akbar
I just want to say Thank you! for that tool!
Hello Akbar,
I like your subtitles translation tool. But I have one issue that makes the translations problematic.
I am translating to german and we have characters like ä, ü, ö, and ß.
The translation comes in well, and I can save it properly.
But if I save and reedit the file again, all these characters get changed to “??”.
Otherwise I am quite happy with the results.
Thank you so much.
Hi Dizzy,
Thanks for your feedback and I’m glad that you are finding this of help.
Regarding your issue, it appears to be a file encoding issue. What you say that “reedit the file again”, what editor you are using to edit the file. Is it that you are upload the file again to my website and editing it or using editor of your choice? If it’s related to my website, I can definitely look into why the encoding is being removed/changed.
Regards, Akbar
Save As function is not working
Sorry for the trouble and late response. Are you still having this issue – as I can’t reproduce this issue. Did you get any error when this happen? If you can share a bit more detail about the flow/error, I can see what went wrong there and how to best fix it.
Regards, Akbar
Hi. I often translate from English to Turkish. But I have an issue. The meaning of the “okay” word is “tamam”. So, “Okay my love”, “Tamam aşkım”. Well, your translation says “Tamam aşkım” when I upload the subtitle the definition is “okay” in context. Actually, must be only “Tamam.”
Sorry for the trouble Murat. From your description, I think the problem is with underlying translation service. Just out of curiosity, what translator you using. Did you try to use any other one and see if that helps.
If nothing else works, you can do manual find/replace to fix this yourself. You can do this directly from the Editor interface as explained below:
https://blog.syedgakbar.com/subtitles-translator-help-tips/
Regards, Akbar
Could you add a regular text/document translator as well? I often find ChatGPT just skipping over parts – with the paid subscription as well. I suppose if you did work through the API, it’d behave normally. And you’d also get the option to use other LLM models. So basically the same idea as the sub translator, just with attachments and/or a huge text field. Pricing could work the same. I really love this translator by the way, its working so good. The prices are also super affordable. Really appreciate the love and effort you put into this project, it shows and its amazing.
Hi Peter, thank you for feedback, and nice comments. I’m glad that you are finding this tool mostly of help.
You have a really interesting idea to support just normal text translation. No one has asked this before, mainly because most of these newer LLMs now support this feature natively. Though I can see your concern with skipping part of translation – I have seen this myself during few API calls. I will think about tweaking the interface to support just text based files. Though it’s not as simple as it first sounds, so I may not be able to do this pretty soon.
Regards, Akbar
Hi Akbar,
Can you upload tutorial video how you build this page on youtube just for learning?
Sure, why not. Hopefully will try to get some timeout to explain basics of this some day soon.
Regards, Akbar
When translating from Hebrew to English, what’s the best way to avoid the punctuation (periods, commas, question marks, exclamation points, etc.) being placed at the beginning of the line, like it is in the original Hebrew, instead of at the end like it should be in English?
After translation, change the language alignment to “Right” (above the Translated Text table column). Also, please make sure “Embed Language Bi-Directional codes” is turned ON under “Save Settings”, so that it plays fine your preferred media player.
Regards, Akbar
That doesn’t seem to do the trick. At least using GPT-4o, it ends up with most of the lines having the punctuation in the right place, at the end of the line. However, some lines still have it at the beginning.
Hmmm… that’s odd. I will have to look into this then. Can you please share some sample subtitle file with few lines showing this problem?
Also, did you try DeepL or Gemini. Just trying to find out if if this is a translator issue or the file encoding/alignment issue.
Regards, Akbar
https://www.opensubtitles.org/en/subtitles/10740491/sunstroke-he
It might just be complexities with the language itself, I’m not sure. Other translators also had the same issue
Thanks for the link. Interesting! I think the problem is with the subtitle itself, where it has applied language direction too. Like if you apply Right-to-Left style on the content, you will notice how punctuation parts are reversed, and that’s probably why all these popular translators are failing, because they don’t expect Hebrew text with direction applied.
Actually, I did just try it with the Gemini Pro free, and it did seem to work okay at least for the first 50 lines. DeepL doesn’t have Hebrew as an option. I suppose I can try to do the full translation with Gemini, but I do prefer GPT-4o.
Interesting, I think Gemini translator is just being smart enough that it knows that punctuation are at wrong place and auto fixes these.
I will probably explore few more of these subtitles and if it’s common scenario, I will fix I can fix this properly.
I would also suggest to try Gemini Flash 2.0 which seems to be doing far better results with punctuation and is slightly cheaper too. I’m still working on making it work with GPT-4o by tweaking the prompt to handle this properly.
Regards, Akbar
So is it something I can fix myself with this file without much work? Otherwise I thought I would just go through and fix it manually.
I don’t know Hebrew, so can’t be of much help in determining how much effort it’s to fix this manually. If it’s just the punctuation symbols, moving these around would not be that tricky, but I’m afraid it might be more tricky/complex than just this.
Regards, Akbar
Oh, I see. I just thought when you said “tweaking the prompt to handle this properly” that I could possibly put something in the prompt extension to handle it.
Yeah, I was taking about tweaking the prompt I internally use for the translation batches. Though that said, you as user can also tweak the prompt instruction by setting “Additional instructions” field, but this doesn’t seem to make much difference in this particular translation tweak.
Regards, Akbar
really very fast and correct translation.
Great tool very useful for me.
We kindly request that you add the DeepSeek platform to your services. Thank you.
Thanks James! This is already in my queue. However, they are not currently giving new API keys to the developers. As soon as their services are back up, I will integrate that.
Regards, Akbar
Hello Akbar! I am currenty unable to log in with my google account, tried on different browsers and OS, it always returns with the same error message:
You have an error in your SQL syntax; check the manual that corresponds to your MySQL server version for the right syntax…
Sorry for the trouble Johannes, I’m investigating this issue and will keep you posted.
Regards, Akbar
Can you add support for the deepseek model?
Already working on that. I will keep you posted.
Regards, Akbar
Great! Thank you for all your work!
Hi
Hiw can i download subtitles in extension.en.srt or .id.srt in save all
Because i need in this and subtitles translator downloads in full names like English Indonesian when downloading as save all in zip
When downloading multiple files, the ZIP file is the only option. I will think about adding some UI to let you individually download the files as well.
Regards, Akbar
ok thanks
Hi, can you add DeepSeek V3 model for translation subtitle, i think that model is cheaper than 01-mini. thanks
Thanks for the suggestion. Work is already in progress on this and hopefully will support this soon.
Regards, Akbar
which one is best way to get srt translation from hindi to english? i need 100% accuracy.
I don’t think currently any machine based translator would give you 100% accuracy for any type of content. I have found that Gemini and OpenAI based translators gives pretty good translation with maintaining context (around 80-90% accuracy), but none of these are prefect and in most of the cases you would need minor editing and tweaks to the translated subtitles to make final translation reach a satisfactory or human translation level.
Regards, Akbar
Hello. Many thanks for this brilliant tool. Is there a way to set the source language. I understand the LLM figures it out 90% of the time. However, there are times, particularly if lines are short, when it mistakes the source language and therefore gives a different meaning in the target language.
True, it definitely helps when the source language is provided. For the LLM translators (Gemini, OpenAI) you can select the source language after you click the “Calculate Cost” button. It’s the first option in the configuration with title “From Language”. I hope this helps.
Regards, Akbar
Tengo problemas, eh traducido un subtitulo dos veces y en cierta parte no lo hizo, no me lo traduce completamente, y pague por la traducción.
Sorry Patrick for the trouble. OpenAI models sometime skip the translation for few of line blocks and I’m already working to fix the properly. In the meantime, once the translation is done, you can re-translate a batch of lines to fix any partial omission or misalignment. You can more about it here:
https://blog.syedgakbar.com/subtitles-translator-help-tips/#retranslate-block
PS. I have also initiated a full refund of this bad translation and you should have the money in your translation balance right away.
Regards, Akbar
Please make it automatically select the same language when I select a language the next time I visit the site.
Thanks, neat suggestions! I will definitely try to work on this soon.
Regards,
Akbar
google free translation have a any new upgrade or accuracy
Not that I know but I will not be surprised if Google auto upgrades the underlying model soon given their success with Gemini LLM model.
Regards, Akbar