I had a similar issue, and the solution I found was to use the Botpress integration in Make. I send only the MID of the WhatsApp audio to Make, which avoids the error caused by URLs like this: "https://lookaside.fbsbx.com/whatsapp_business/attachments/?mid=1073950600712166&ext=1731262529&hash=ATsdQDGH5J8iXTMKr8BiWqddt851whlfZBIJ2jRY2wDjyg". With the MID in Make, I simply integrate it with WhatsApp’s media download feature. Once downloaded, it’s converted into a binary-like format. I then pass this format to ChatGPT to convert the audio into text.
In my workflow, the transcribed text is saved to a variable, which I then import into my CRM.