TTS-1 HD

TTS-1 HD launched on 6 November 2023 alongside TTS-1 and offers the same six preset voices through the same speech endpoint. The difference is the trade-off each model makes: TTS-1 is tuned for real-time latency, TTS-1 HD for audio quality. Everything else — the interface, the voice set, the absence of a context window or tool use — is shared. The pricing relationship between the two is unusually clean. TTS-1 costs $15 per million characters, TTS-1 HD costs $30: exactly twice as much, for the same text turned into the same voices at higher fidelity. OpenAI has never published a benchmark quantifying how much better the audio is, so the choice between them is a judgement call rather than a calculation — a rare case where the vendor prices a difference it does not measure publicly. Both models remain on the price list nearly three years after launch, unchanged in price, while the rest of OpenAI's audio offering moved to the multimodal line. For a developer generating pre-rendered audio — audiobook chapters, recorded prompts, anything not spoken live — HD is still the plainest option: no instructions to write, no reasoning tokens to pay for, one number per character.

▸Videos

No videos yet.

▸Reviews