Speech 2 HD key stats
- Latency (measured)
- 357 ms1
- Vapi streaming benchmark (50 trials per model) (checked 2026-06-11) Median of 50 sequential live streaming trials, June 2026; includes network RTT from the benchmark machine.
- arxiv.org/abs/2505.07916 (checked 2026-06-11) MiniMax-Speech technical report (the Speech-02 architecture) lists 32 languages.
- platform.minimax.io/docs/guides/pricing-paygo (checked 2026-06-11) T2A pay-as-you-go: speech-02-hd at $100 per 1M characters.
- minimax.io/news/speech-02-series (checked 2026-06-11) Series launch post dated April 2, 2025; rollout coverage ran through May 2025.
Background
Speech 2 HD (speech-02-hd in the API) is the high fidelity tier of the MiniMax Speech 2 generation, launched in April 2025 for voiceover and audiobook work where rhythm consistency and clarity matter most. The generation behind it, documented in the MiniMax-Speech technical report, pairs an autoregressive Transformer with a learnable speaker encoder that clones a voice from roughly ten seconds of reference audio with no transcript, across 32 languages. Speech 2 HD topped the Artificial Analysis Speech Arena ELO rankings on release, ahead of OpenAI and ElevenLabs.
Sources: minimax.io, arxiv.org
At a glance
The quality tier of the Speech 2 pair, against Speech 2 Turbo as the realtime tier. In our 50 trial streaming benchmark it returned first audio in a median of 357 ms including network time, close behind its Turbo sibling despite the fidelity focus.
Sources: platform.minimax.io
Position in the rankings
Standings as of Jun 13, 2026, 01:14 UTC
Frequently asked questions
- How is Speech 2 HD tested on the Humanness Index™?
- Listeners hear Speech 2 HD against another model in a blind head to head round, both voices reading the same customer support prompt from the same cloned source voice, and they pick whichever sounds more human. Its Humanness score derives purely from those votes.
- How does Speech 2 HD differ from Speech 2 Turbo?
- Same generation, different tuning. Speech 2 HD targets high fidelity output for voiceovers and audiobooks and bills at $100 per 1M characters; Speech 2 Turbo targets realtime interaction at $60 per 1M. In our 50 trial benchmark they measured 357 ms and 315 ms median time to first audio respectively.
How human does your model really sound?
The benchmark is open source. Suggest a model, read the methodology, or ask us to put your voice in the arena.
