Anyone else experiencing this right now? Had a working TTS pipeline, and today ~1 out of 10 generations has a word or two cut off at the end. Did something change in the API?
The entire ChatGPT framework is becoming FURTHER eroded with each update. I’m basically stepping back and using other AI models like Grok, Manus, Deepseek R1m, Qwen, Perplexity, and Claude.
+1 same here using gpt-4o-mini-realtime-preview-2024-12-17
sometime of the last audio bits are missing, but transcript is complete.