AI translation engine management (available on the Premium plan) allows you:
More control over the quality of the AI Translations, by using your own API key for one of the supported translation engines.
Allows unlimited Products translations (for merchants exceeding the 15,000 products limit on the Premium plan)
The translation engines we support are:
OpenAI
DeepL
DeepSeek
Google Translate
How to add your own API key?
To add your own API key, go to the Settings > AI Translation Settings page
Choose the desired engine, then add our API key, and click "Verify key" button on the right:
If you get a "Verified" success message, you are good to go - future AI translations will use the engine you have selected. Otherwise, if the verificaiton fails, please make sure:
Your key has permissions
There are no billing issues with your account at the chosen platform.
Engine-specific settings
Some of the engines have specific settings that you can use for further customization of the transalations.
OpenAI additional settings
You can create your OpenAI account and API key on https://platform.openai.com/login
Using OpenAI API key allows you to:
ChatGPT model - Choose the ChatGPT model you want to use for the translations
Tone of voice - default, friendly and professional
Additional translation instructions - an "additional prompt" that will be used for every translation
Note that longer instructions consume more tokens and may affect your OpenAI billing.
DeepL additional settings
You may create DeepL account at https://www.deepl.com/en/signup
Adding your DeepL API key allows you to configure:
DeepL formality - the formality setting controls whether translations are generated in a more formal or more informal tone for supported languages.
Example:German formal: uses “Sie”
German informal: uses “du”
If the target language supports it, DeepL adjusts the grammatical politeness level. If not, the setting is ignored.
Model type - the Model Type setting controls whether translations prioritize speed or quality.
latency optimized → Faster translations, slightly less refined.
quality optimized → Higher linguistic quality, slightly slower.
prefer quality optimized → Uses the quality model when available, otherwise falls back automatically.
Selecting the higher quality model, will not result in higher cost, as DeepL charges depend on number of translations and subscriptions.
Glossary ID - if you have created a glossary in DeepL you may add it to the settings, so that it is taken into account during the AI translations.
If you add your DeepL glossary ID, the TLab Glossary rules are ignored.
In order for the DeepL glossary to work successful, you need to make sure that Source language detection setting is set to Use store default language:
As DeepL glossary works only if it knows exactly which are the language pairs.
DeepSeek additional settings
You can create DeepSeek account at https://platform.deepseek.com/sign_in
Adding DeepSeek API key allows you to choose:
Tone of voice - default, friendly or professional
Additional translation instructions - work as an "additional prompt" that will be used for every translation
Note that longer instructions consume more tokens and may affect your billing.
FAQ
Question: Which is the best translation engine - OpenAI, DeepL, DeepSeek, or Google Translate?
Answer: There is no universally “best” translation engine.
Quality and cost vary depending on:
The language pair (e.g., English → German vs English → Thai)
The type of content (marketing, legal, technical, product data)
The need for terminology control (glossaries, brand tone)
Speed vs refinement requirements
Your budget and API pricing model
Some engines perform better for European languages, others may handle Asian languages differently. Pricing structures also differ - some charge strictly per character, others per token or usage tier.
The most reliable approach is to test multiple engines with a few of your products (not all the content), then compare quality, consistency, and cost for your specific use case.
Question: How much would the translations cost be with each engine?
Answer: Pricing structures differs - some translation engines charge strictly per character, others per token or usage tier. It's best to refer to the selected engine's pricing page for reference.
Since it's impossible to calculate the exact amount of words/characters a given store has, the best way to get an idea is to translate 10 products, for example, then check your billing at the selected platform (note that some of them take a few hours to update your usage statistics)