These watermarks consist of special #Unicode characters (primarily the Narrow No-Break #Space) that look identical to regular spaces but have different #ASCII - codes. In our testing we did not observe watermarks in older models such as #GPT-4o.
The newer GPT-o3 and GPT-o4 mini models appear to be embedding special character watermarks in generated text. However, removing these watermarks is relatively simple, making this seem more like a short-term measure than a long-term solution
text2field-recordings - a live project which can be booked! texts are played on a keyboard: every letter triggers a specific fieldrecording that was previously filed in a thematic sound container. the text is the composition. the performer controls only the length and velocity. #text #fieldrecording #bookingbase.at/t2f/
