DeepL makes translated speech generally available and adds a translation watermark option

Two recent items from DeepL’s developer changelog matter for translation professionals.

Speech-to-speech — translated speech output — has moved out of closed beta and is now generally available through the company’s Voice products. DeepL first announced its Voice-to-Voice suite in April 2026, positioning it for live meetings, in-person conversations and customer-facing use via API.

Separately, the document translation API now accepts an enable_watermark parameter: when set, translated Word and PDF files carry a visible “Translated by DeepL” mark.

The watermark is a small change with an interesting implication: machine-translated documents can now be labeled as such at the source. For professional translators, provenance labeling cuts both ways — it makes unedited machine output easier to identify, and it sharpens the distinction between raw MT and human-reviewed translation, which is where professional value increasingly concentrates. Post-editing, terminology control and accountability for high-stakes text remain outside what the API automates.