{
  "id": 12976060,
  "title": "OpenAI silences familiar voices",
  "url": "https://urgent.news/2026/10/08/openai-silences-familiar-voices",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-08T22:50:22.000Z",
  "source": {
    "name": "Six Colors",
    "slug": "six-colors",
    "url": "https://sixcolors.com/post/2026/10/openai-silences-familiar-voices/"
  },
  "original_language": "en",
  "account": "OpenAI has announced that it will be discontinuing the use of certain legacy audio models, including tts-1, tts-hd, and gpt-4o-mini-tts. These models will be shut off on January 6, 2027, as part of the company's continuous upgrade process. OpenAI's text-to-speech voice is currently used for the Six Colors Audio Newsletter, and the producer of the newsletter had to make significant changes to his workflow after receiving an email from OpenAI about the deprecation.\n\nThe producer of Six Colors Audio Newsletter had chosen a less advanced voice from OpenAI's offerings for the newsletter due to its reliable performance. This decision was inspired by another service called Listen Later, which turns saved articles into a podcast using a now-deprecated OpenAI TTS voice. However, the closure of these voices has caused disappointment for both the newsletter producer and the users of Listen Later, as their brains have grown accustomed to the specific patterns and delivery of those voices.\n\nThe closure of these voices raises concerns about accessibility, particularly for visually impaired users who rely on Listen Later's functionality. The producer of Six Colors Audio Newsletter has been investigating alternative options to replace the deprecated models, ultimately opting for the newer ElevenLabs v4 engine. While this alternative comes with a monthly cost, it has proven to produce satisfactory results.\n\nTo ensure accuracy and prevent AI errors, the producer has implemented a workflow that tests every audio file generated by the API, comparing the output to the original text. This process, while time-consuming, allows the newsletter to utilize more sophisticated models while minimizing the risk of omissions or hallucinations. The newsletter currently runs both the OpenAI voice and the ElevenLabs alternative simultaneously for testing purposes, with the hope that users will appreciate the new voices.",
  "summary": "I had to listen to a bunch of TTS voices to make my final choices. An audio newsletter producer’s job is never done. I got a weird email the other day from OpenAI that ended up causing a great deal of work for me: As part of our continuous upgrade process, we are deprecating the following legacy audio models: tts-1, tts-hd, gpt-4o-mini-tts. …",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}