When calling Cartesia's Clone Voice API, send the clip as multipart form data under 16 MB in a supported format, and always include name and a valid language code. Set accent metadata up front if you plan to add accents later. Use the access field instead of the deprecated access[type] form field.

Context: Official Cartesia docs (Clone Voice API reference): documents the request constraints agents hit. The clip upload is multipart form data with a 16 MB maximum and a limited format list (flac, mp3, mpeg, mpga, oga, ogg, wav, webm). Name and language are required fields; the language must be one of the supported codes. The accent field is only metadata for labeling and is a prerequisite for the Add Voice Accents flow. The deprecated access[type] form field is still accepted, but the plain access field with private or public is preferred.