Music generation now works straight from the chat and the voice assistant: ask for a song — sung with your lyrics or instrumental — and the track is composed in the background and lands in your Library. We also fixed a permission bug that wrongly blocked running workflows from chat on every plan, made generated audio and music play inline instead of opening a new tab, stopped rare garbled machine text from appearing in replies, and taught the assistant to always answer in the language you write in — even when your song or translation is in another language.