ProZ.com Translation Forums

I am building a AI translation website and CAT tool | Download

CafeTran Trainer — Sat, 14 Mar 2026 09:40:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: CafeTran Trainer
Post title: Download

Is it an online-only service?

I am building a AI translation website and CAT tool | Nice job!

Li-Hsiang Hsu — Sat, 14 Mar 2026 06:58:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: Li-Hsiang Hsu
Post title: Nice job!

Great for translators not using Trados 2024.

I am building a AI translation website and CAT tool | Batch Translate

Zijia He — Sat, 14 Mar 2026 01:52:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: Zijia He
Post title: Batch Translate

Hi Michael, I am quite interested in the 'Batch Translate' part of your projects.

My idea is for AI to read the context first and generate translation for empty segments.

The result of AI translation for only a single segment is not good enough as context is not considered

[quote]Michael Beijer wrote:

[quote]Zijia He wrote:

Hi Guys,

I am building tools to better use AI translation. I am a translator and PM, not good at coding, so using AI coding to realize my thought on improving translation quality and tools with AI.

Website: wordplay.icu

Suggestion and chat are more than welcomed!!!

Sid [/quote]

Hi Sid,

Interesting stuff, thanks for sharing. I am in pretty much the same boat myself. I am a translator with only very limited coding experience; however, the latest AI coding tools have allowed me to realise things I have been dreaming about for years. It's absolutely insane what can be achieved with Claude Code these days. I have been working on a similar project called Supervertaler.

Several of my colleagues are also currently working on their own tools, either privately or publicly, and I always like to see how others approach things from their perspectives.

Michael [/quote]

I am building a AI translation website and CAT tool | True

CafeTran Trainer — Fri, 13 Mar 2026 20:01:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: CafeTran Trainer
Post title: True

[quote]Jennifer Levey wrote:

Fair competition between peers and colleagues is one of the most healthy activities we can engage in through the Proz platform. [/quote]

I agree!

I am building a AI translation website and CAT tool | Interesting stuff, thanks for sharing!

Michael Beijer — Fri, 13 Mar 2026 14:06:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: Michael Beijer
Post title: Interesting stuff, thanks for sharing!

[quote]Zijia He wrote:

Hi Guys,

I am building tools to better use AI translation. I am a translator and PM, not good at coding, so using AI coding to realize my thought on improving translation quality and tools with AI.

Website: wordplay.icu

Suggestion and chat are more than welcomed!!!

Sid [/quote]

Hi Sid,

Interesting stuff, thanks for sharing. I am in pretty much the same boat myself. I am a translator with only very limited coding experience; however, the latest AI coding tools have allowed me to realise things I have been dreaming about for years. It's absolutely insane what can be achieved with Claude Code these days. I have been working on a similar project called Supervertaler.

Several of my colleagues are also currently working on their own tools, either privately or publicly, and I always like to see how others approach things from their perspectives.

Michael

I am building a AI translation website and CAT tool | Why stifle a colleague's initiative?

Jennifer Levey — Fri, 13 Mar 2026 13:05:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: Jennifer Levey
Post title: Why stifle a colleague's initiative?

[quote]CafeTran Trainer wrote:
[quote]Zijia He wrote:
I am building tools to better use AI translation. I am a translator and PM , not good at coding, so using AI coding to realize my thought on improving translation quality and tools with AI.[/quote]
You can save a lot of time by working with the code of Supervertaler ...

See: [url removed]
[/quote]

I am no fan of any form of exploitation of AI in translation.

That said, I find it absolutely inappropriate to suggest that the OP should seek to 'save a lot of time' by cribbing the ideas and code of another Proz.com member. That only stifles innovation, diminishes creativity in our profession, and impedes personal/professional development, and self-learning, of our colleague from China. And at the same time it suggests - without justification - that Supervertaler and its developer have already got everything fine-tuned to perfection.

Fair competition between peers and colleagues is one of the most healthy activities we can engage in through the Proz platform. Don't stifle it just 'to save time'.
JL

Translating long documents: where does consistency break? 5–7 min survey (shaping a new tool)

Roman Makarov — Fri, 13 Mar 2026 12:52:00 +0000

Forum: Machine Translation (MT)
Topic: Translating long documents: where does consistency break? 5–7 min survey (shaping a new tool)
Poster: Roman Makarov

Hi everyone 👋

I’m building a tool for translating/editing long documents and books, where common pain points are:

terminology drifts after 50+ pages
glossaries exist but aren’t enforced consistently
“replace everywhere” breaks context
reviewing/fixing takes longer than the first pass

I’d really value your experience. This is a short survey (5–7 minutes) and your answers will directly influence what gets built (glossary workflow, consistency checks, review mode, batch edits).

👉 Take the survey

Note: you can answer without any confidential data—high-level examples are enough.

Thank you! If you want, reply here with your #1 pain point on long projects. 🙏

I am building a AI translation website and CAT tool | Already done

CafeTran Trainer — Fri, 13 Mar 2026 06:46:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: CafeTran Trainer
Post title: Already done

[quote]Zijia He wrote:

Hi Guys,

I am building tools to better use AI translation. I am a translator and PM , not good at coding, so using AI coding to realize my thought on improving translation quality and tools with AI.

Website: wordplay.icu

Suggestion and chat are more than welcomed!!!

Sid [/quote]

You can save a lot of time by working with the code of Supervertaler ...

See: [url removed]

[Edited at 2026-03-13 06:48 GMT]

I am building a AI translation website and CAT tool

Zijia He — Thu, 12 Mar 2026 12:26:00 +0000

Forum: Machine Translation (MT)
Topic: I am building a AI translation website and CAT tool
Poster: Zijia He

Hi Guys,

I am building tools to better use AI translation. I am a translator and PM , not good at coding, so using AI coding to realize my thought on improving translation quality and tools with AI.

Website: wordplay.icu

Suggestion and chat are more than welcomed!!!

Sid

Recommended tools for machine translation tools | Updated Recommendation for Japanese to English Machine Translation

Ronika Kashyap — Mon, 28 Jul 2025 12:35:00 +0000

Forum: Machine Translation (MT)
Topic: Recommended tools for machine translation tools
Poster: Ronika Kashyap
Post title: Updated Recommendation for Japanese to English Machine Translation

I know this thread is 4 years old, but it still shows up in searches, so here’s a more up-to-date suggestion. If you’re looking for a better Japanese>English machine translation tool than ATLAS, try Pairaphrase. It’s more flexible with file formats, keeps formatting intact, and gives you editable, context-aware translations. Here’s a helpful overview:

[url removed]

Issues with DeepL web | Maybe too much 'reflected radiation' a.k.a. gorging on own 'recycling'?

Daryo — Tue, 22 Jul 2025 06:17:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Daryo
Post title: Maybe too much 'reflected radiation' a.k.a. gorging on own 'recycling'?

'DeepL has become almost unusable'?

see:

Low-background Steel (pre-AI)

Sources of data that haven’t been contaminated by AI-created content. Low-background Steel (and lead) is a type of metal uncontaminated by radioactive isotopes from nuclear testing. That steel and lead is usually recovered from ships that sunk before the Trinity Test in 1945. This blog is about uncontaminated content that I'm terming "Low-background Steel". The idea is to point to sources of text, images and video that were created prior to the explosion of AI-generated content that occurred in 2022.

[url removed]

All very predictable, but since when 'predictable' was a factor when making a quick buck is at stake?

Issues with DeepL web | Barely usable anymore.

Christian Schneider — Sun, 13 Jul 2025 18:49:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Christian Schneider
Post title: Barely usable anymore.

At the moment DeepL has become almost unusable both the next generation language model as well as the the conventional model.
Way back then I could modify the output to form exactily the sentences that I wanted. This is no longer possible. I cnceled my annal subscription for now and hope it will change for the better.

Issues with DeepL web | I understand

Matthias Brombach — Wed, 28 May 2025 18:29:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Matthias Brombach
Post title: I understand

[quote]Baran Keki wrote:
In a CAT you won't be able to edit the translations and get terminology suggestions the way you do on DeepL web or desktop. [/quote]

Good point and my friend, who also uses DeepL for professional purposes, suggests to wait until DeepL soon will issue the "Holy Ghost" edition of its services, with perfect consistent results based on true knowledge and experience DeepL has gained by true artificial intelligence. No need to fiddle with annoying terms throughout all segments of a CAT project. All results will be the truth and nothing but the truth, no matter whether the source text has been produced by a non native speaker, a term plumber or one of the numerous patent translators with a degree in books for children.

Issues with DeepL web | Useless

Baran Keki — Wed, 28 May 2025 16:11:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Baran Keki
Post title: Useless

[quote]Matthias Brombach wrote:

...the Deepl API in your CAT? Not as if I use DeepL at all, I mean, who does, but a friend of mine told me that you can easily integrate DeepL in your CAT. [/quote]
In a CAT you won't be able to edit the translations and get terminology suggestions the way you do on DeepL web or desktop. I've sold my soul to the devil long time ago.
It's interesting that I seem to be the only person suffering from this issue.

Issues with DeepL web | Next-Gen vs Classic

Jean Dimitriadis — Wed, 28 May 2025 14:26:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Jean Dimitriadis
Post title: Next-Gen vs Classic

I've found this:

[url removed]

In particular, it says that: The use of our own LLMs for the next-gen language model provides improved quality for the translation of your input texts, especially texts that are longer.

I am guessing that for longer texts, past a certain point the system somehow switches to the next-gen model? And since switching between models triggers a retranslation, that would explain the loss of the edited translation.

Clearly, this switch and content loss should not be happening, especially without a way to confidently undo the changes.

I highly recommend you open a support ticket for this, as like you say, it is disrupting the tool's usability.

Issues with DeepL web | Why not using...

Matthias Brombach — Wed, 28 May 2025 13:55:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Matthias Brombach
Post title: Why not using...

...the Deepl API in your CAT? Not as if I use DeepL at all, I mean, who does, but a friend of mine told me that you can easily integrate DeepL in your CAT.

Issues with DeepL web

Baran Keki — Wed, 28 May 2025 08:33:00 +0000

Forum: Machine Translation (MT)
Topic: Issues with DeepL web
Poster: Baran Keki

I'm using the web version of DeepL Pro (supposedly the paid version) as I find it more convenient than the desktop version, and I've been having the following annoying issues with it for the last month or so:
1- There seems to be two options in the dropdown menu under the options that says 'Classic' and 'Next Gen'. I always use Classic as it provides more phrase suggestions compared to 'Next Gen' (I don't know what is so 'next' about it as it gives less than a fifth of what Classic gives suggestion-wise and they are usually wrong ones). It works okay for a while, but all of a sudden it changes the setting from Classic to Next Gen, and it screws up the translation that you've done up to that point, and everything goes back to default. This happens especially when you've pasted a number of sentences (say 3 paragraphs containing 8 or 9 sentences) on the source pane, and usually when you're working on the 5th or 6th sentence. All of the work you've done, manual edits etc., gets lost and you get a lot less phrase, term suggestions going forward. Changing it from Next Gen to Classic won't matter from that point on, as it will forever go back to Next Gen. You'll have to close that tab and open a new one and start the process all over again.
2- There used to be undo/redo buttons that you could always use, now they appear to be available haphazardly. The undo button gets grayed out when you're translating multiple sentences. Sometimes it's miraculously available, but most of the time it isn't. It's extremely frustrating.
3- Almost 90 percent of the time you don't get any suggestions for the first word of a sentence. When you click on it, you get nothing. I don't remember this happening before.
4- It seems that when you leave the translation idle for an extended period of time (anywhere between 5 to 15 minutes) and then click on a word, everything reverts to default, again all your previous work gets screwed up (and needless to say it goes from Classic to Next Gen again). And again, I don't remember this happening before.

It's almost impossible to work efficiently with it on a large project. I've installed the desktop version recently and had the same issues again (it reverted to the default translation after clicking on a word, following a 15 minute inactivity).
Is anyone having the same problems? What should I do to make it run smoothly as before? I'm kind of reluctant to clear the history or cache of my browser. But if needs must, I'll do it.

My experience is not for sale | If words could kill..

Baran Keki — Mon, 12 May 2025 10:41:00 +0000

Forum: Machine Translation (MT)
Topic: My experience is not for sale
Poster: Baran Keki
Post title: If words could kill..

You were right to ask for more money for 'doing hits'. That said, '9500 hits' is an unconscionably high number. Almost at an industrial scale.

My experience is not for sale | So you...

WolfgangS — Mon, 12 May 2025 09:35:00 +0000

Forum: Machine Translation (MT)
Topic: My experience is not for sale
Poster: WolfgangS
Post title: So you...

[quote]Tom in London wrote:

They wrote:

Hello Thomas,

We hope you are doing well.

This is XXX Vendor coordinator at XXX company, nice to e-meet you 😊

We are checking your availability for a new task, which we would like to assist us. Please check the following details:

Language pair: Italian to English
Subject matter: General
Service: Annotation
Volume: 9500 hit, the hit is between (20 to 200 word)
Start date: the task is already in progress

If you are interested and meet the required needed criteria, please provide the below details:

Your rate per hit for annotation.
Your CV

We look forward to your response.

My response:

Dear XXX

If you want me to use my bilingual ITA/ENG professional expertise and experience, acquired over many years, to help you to improve your AI machine, it will be very expensive. You would not be able to afford me.

Regards

Tom [/quote]

even bothered to answer???

[Edited at 2025-05-12 09:43 GMT]

My experience is not for sale

Tom in London — Mon, 12 May 2025 09:27:00 +0000

Forum: Machine Translation (MT)
Topic: My experience is not for sale
Poster: Tom in London

They wrote:

Hello Thomas,

We hope you are doing well.

This is XXX Vendor coordinator at XXX company, nice to e-meet you 😊

We are checking your availability for a new task, which we would like to assist us. Please check the following details:

Language pair: Italian to English
Subject matter: General
Service: Annotation
Volume: 9500 hit, the hit is between (20 to 200 word)
Start date: the task is already in progress

If you are interested and meet the required needed criteria, please provide the below details:

Your rate per hit for annotation.
Your CV

We look forward to your response.

My response:

Dear XXX

If you want me to use my bilingual ITA/ENG professional expertise and experience, acquired over many years, to help you to improve your AI machine, it will be very expensive. You would not be able to afford me.

Regards

Tom

Does ChatGPT add watermarks? | I can't see any need for any watermarks ...

Daryo — Wed, 07 May 2025 07:55:00 +0000

Forum: Machine Translation (MT)
Topic: Does ChatGPT add watermarks?
Poster: Daryo
Post title: I can't see any need for any watermarks ...

Corrected

Does ChatGPT add watermarks? | I can't any need for any watermarks ...

Daryo — Wed, 07 May 2025 07:50:00 +0000

Forum: Machine Translation (MT)
Topic: Does ChatGPT add watermarks?
Poster: Daryo
Post title: I can't any need for any watermarks ...

Its insipid bland fake neutral 'writing style' makes the whole of CheatGPT's output an unmistakable watermark / 'proof of origin'.

Does ChatGPT add watermarks? | CheatGPT

Tom in London — Wed, 07 May 2025 07:46:00 +0000

Forum: Machine Translation (MT)
Topic: Does ChatGPT add watermarks?
Poster: Tom in London
Post title: CheatGPT

[quote]Samuel Murray wrote:

Hello everyone

[/quote]

Does ChatGPT add watermarks? | ChatGPT

Samuel Murray — Wed, 07 May 2025 06:27:00 +0000

Forum: Machine Translation (MT)
Topic: Does ChatGPT add watermarks?
Poster: Samuel Murray
Post title: ChatGPT

ChatGPT says that if its output contains hidden characters, it is due to a quirk in your browser.

Does ChatGPT add watermarks?

Samuel Murray — Wed, 07 May 2025 06:26:00 +0000

Forum: Machine Translation (MT)
Topic: Does ChatGPT add watermarks?
Poster: Samuel Murray

Hello everyone

TikTok is awash with people telling you how to remove ChatGPT watermarks. According to the videos, ChatGPT inserts invisible characters such as no-break spaces and zero-width spaces in random places. Some videos state that the watermarking only applies to certain engines. These videos often link to the character viewer soscisurvey.de/tools/view-chairs.php (now also available on GitHub).

Have any of you seen these watermarks? Is this a real thing? ChatGPT itself claims that it does not use watermarking, but the answer that ChatGPT gave me refers to lexicon-based watermarking systems.

It feels like it would be unlikely since it would be relatively straight-forward to include a watermark remover to any AI tool.

Samuel

[Edited at 2025-05-07 06:27 GMT]

Is GT4T still supported/maintained? | Thank you

Elena Aleksandrova — Thu, 10 Apr 2025 14:09:00 +0000

Forum: Machine Translation (MT)
Topic: Is GT4T still supported/maintained?
Poster: Elena Aleksandrova
Post title: Thank you

[quote]Stepan Konev wrote:

Probably you can contact the GT4T dev directly. Here is their profile: [url removed] [/quote]

Thank you :) Already sent.

Is GT4T still supported/maintained? | GT4T Developer

Stepan Konev — Thu, 10 Apr 2025 14:05:00 +0000

Forum: Machine Translation (MT)
Topic: Is GT4T still supported/maintained?
Poster: Stepan Konev
Post title: GT4T Developer

Probably you can contact the GT4T dev directly. Here is their profile: [url removed]

Is GT4T still supported/maintained?

Elena Aleksandrova — Thu, 10 Apr 2025 12:49:00 +0000

Forum: Machine Translation (MT)
Topic: Is GT4T still supported/maintained?
Poster: Elena Aleksandrova

There is an issue with GT4T not picking the source since yesterday morning. I've sent 3 emails to 2 different email addresses (from the app and from the website), and no answer. Recently, there was a problem, and it was solved within hours.

Do you know if it is still supported/maintained? I'd like to know should I expect an answer and fixing the problem or should I get another MT.

Thank you.

Experience with DeepL Voice for Meetings

FranziskaSo — Tue, 25 Mar 2025 14:38:00 +0000

Forum: Machine Translation (MT)
Topic: Experience with DeepL Voice for Meetings
Poster: FranziskaSo

Hello everyone,

since the product is so new, I haven't been able to find any honest reviews of DeepL Voice for Meetings.

The company I work at is considering implementing it, but I would like to find out first if anyone has already tested it and what their experience was like.

Thank you in advance!