Frequently asked questions

What is Trad-AI?

Trad-AI is a lightweight translation tool that allows you to translate documents using your own OpenAI API key. After translation, you can download both the translated file and a TMX translation memory for use in CAT tools.

Does Trad-AI have its own model?

No. Trad-AI does not host, train, or provide custom models. All translations are generated by OpenAI models through your API key.

Why use my own API key?

Using your own key ensures:

• you stay in full control of usage costs,
• you know exactly which model is being used,
• your data is processed according to OpenAI’s security rules.

How do I translate a document?

1. Select Source and Target languages.
2. (Optional) Select Tone and Domain.
3. Upload a supported file or enter text.
4. Click Translate.
5. Download the translated file and/or the TMX.
6. Click Clear to start a new translation session.

Which file formats are supported?

DOCX, PPTX, PDF, and XLSX, as well as plain text input.

Do I need to format my document before uploading?

Clean, well-structured documents generally translate more predictably. If your source contains broken formatting or text embedded in images, the result may vary.

What is a TMX file?

TMX (Translation Memory eXchange) is a standard format for storing aligned source–target text segments.

How do I use the TMX file?

Import it into your CAT tool (such as Trados, memoQ, Smartcat, Wordfast, etc.) to:

• apply pre-translation based on the memory,
• improve consistency,
• speed up post-editing.

Do I need to use CAT tools to work with Trad-AI?

No. CAT tools are optional. You can simply download the translated file if that suits your workflow.

What are token limits?

OpenAI models process text in tokens (small chunks of text). Each model has a maximum number of tokens it can handle per request. This limit comes from OpenAI, not from Trad-AI.

How large can my file be?

This depends on the OpenAI model and your API settings. As a general approximation: 1 token ≈ 3–4 characters of English text. Because the translated text is usually similar in length to the source, the token limit effectively sets how much text can be translated at once.

What happens if my file is too large?

If a document exceeds the model’s token limit, OpenAI will return an error such as “context length exceeded”. In that case, the document will need to be split into smaller parts before translating.

Does Trad-AI increase or bypass token limits?

No. Trad-AI does not override or modify OpenAI’s limits.

Is there a demo?

Yes. The public demo lets you test the translation workspace without creating an account.

What are the demo limits?

The demo supports translations of up to 150 characters (including spaces) for both text input and file uploads.

How do I translate longer documents?

To translate full documents and unlock TMX export, select a paid plan.

How am I charged for translation usage?

Translation usage is billed directly by OpenAI, through your own API key. Trad-AI does not charge per character or per request.

What do Trad-AI plans unlock?

Plans provide access to:

• full document translation beyond demo limits,
• downloading TMX files,
• continuous use without input restrictions.

Do you store payment details?

No. Payments are processed securely through third-party billing infrastructure.

Do you store my OpenAI API key?

No. The key is used only to process your translation request and is not stored long-term.

Do you store my documents or translation texts?

No. Content is processed only to obtain the translation result and is not retained.

Do you train models on user data?

No. Trad-AI does not train or fine-tune models using your content.

Who does the actual text processing?

OpenAI processes the text according to its own security and data policies.

Why did I receive a “context length exceeded” error?

Your document is larger than the selected model can process in one request. Split the file into smaller sections and translate them separately.

Why do some PDFs translate less accurately?

If the text in the PDF was not real text (for example, scanned as an image), the extracted text may be fragmented or missing. Converting the PDF to DOCX may improve results.