I’m asking OpenAI to reconsider the January 6, 2027 shutdown of tts-1 and tts-1-hd, and to take the human connection to these voices into account.
I run Listen Later, which turns articles and newsletters into audio. We’ve used these models for nearly three years. They continue to work extremely well for our customers, and hundreds of people have spent years listening to their chosen voice.
Over time, a familiar narrator becomes part of someone’s routine: how they follow the news, enjoy an essay, or catch up on something they care about. The sound, rhythm, and delivery become familiar. There is a real human connection there.
Many of our customers are visually impaired. For many of them, these models were their first experience of article narration at this level of quality. OpenAI built something that made a meaningful difference in their lives. I hope that achievement, and the people who now rely on it, will carry real weight in this decision.
A newer model can be technically excellent and still fail to replace the particular voice someone has listened to for years. Familiarity is itself part of the quality of that experience. Better capabilities do not automatically preserve it.
An extended migration period would help developers, but it would not resolve the underlying loss if those familiar voices disappear. We can change an integration. We cannot recreate years of familiarity for our listeners.
Before finalizing the shutdown, would OpenAI consider speaking directly with affected users, including blind and visually impaired listeners? Their experiences should help inform what an acceptable transition looks like.
My first request is to reconsider the shutdown and keep these models available. If that is not possible, please consider:
- Continuing access for existing applications and customers through a legacy offering that smaller services can realistically use.
- Preserving these familiar voices and their narration characteristics in a supported successor, with affected listeners involved in evaluating the result.
- Enabling another provider to continue serving them through licensing or a hosting partnership, or exploring an open-source release of the models and the components needed to preserve their voices.
I appreciate that maintaining older models involves real costs and tradeoffs. I’m asking that the value of continuity to the people using them be included in that calculation.
These voices are one of OpenAI’s early successes that became part of people’s everyday lives. Please help us preserve something your technology made possible—and that our customers still value.