Yo beautiful people 👋 Does anyone here use the stack Deepgram as SST + GPT-4.1 mini (OpenAI) as model + MiniMax as TTS? I’ve run into a pretty weird pronunciation/formatting issue and I’d love to hear how you’ve dealt with it. Specifically: 🔹 Dates like “2025” are being spoken as “two zero two five” instead of “twenty twenty-five” or “two thousand twenty-five”. 🔹 Times like “8 AM” are being synthesized in a broken-up way — e.g., “eight a m” (separate phonemes) instead of a natural “eight am”. Has anybody encountered this with Deepgram’s TTS or MiniMax speech? What config options, text-preprocessing, prompt formatting, or normalization tricks have you used to force correct pronunciation of dates/times/numbers? Would really appreciate any tips, best practices, or workarounds folks have found. Thanks in advance 🙏