The audio is good, but transcribe text is not. Repsonse from LLM are chunked / split / bad https://cdn.discordapp.com/attachments/1211484481848873020/1493815525866733701/Screenshot_2026-04-15_at_11.28.54.png?ex=69e0579f&is=69df061f&hm=cc3f01997052236b0a440245ed4b058753d1d1c380b9eca1bd9cc7afd37b8652&