MAI-Voice-2.1 English voices with call-centre or audiobook styles
customer_call_center and audiobook styles sit on en-US Grant, Harper, Sage and en-GB Emily, Harry. Iris, Jasper, Dhruv and Priya are neutral only.
Per the Microsoft Learn MAI voices page read on 2026-10-04, the customer_call_center and audiobook styles are on en-US Grant, Harper and Sage and on en-GB Emily and Harry. Iris, Jasper, Dhruv and Priya offer neutral only, and Ethan, Olivia and Isla list emotion styles instead of roles.
The English list
| Voice | Locale | Style type |
|---|---|---|
| Ethan | en-US | 19 emotion styles |
| Olivia | en-US | 19 emotion styles |
| Grant | en-US | 6 role styles (agent, audiobook, customer_call_center, educational, narrator, neutral) |
| Sage | en-US | Same 6 role styles |
| Harper | en-US | 20 styles mixing roles and emotions |
| Iris | en-US | Neutral only |
| Jasper | en-US | Neutral only |
| Emily | en-GB | Role styles and some emotions |
| Harry | en-GB | Role styles and some emotions |
| Isla | en-AU | 19 emotion styles |
| Dhruv | en-IN | Neutral only |
| Priya | en-IN | Neutral only |
What this means for an IVR or a book
For an IVR line pick a voice that lists customer_call_center. For long-form narration, pick one that lists audiobook or narrator. Remember the out-of-scope note on the page: the long-form model prioritises naturalness over latency, so it is not the choice for live conversation.
The voice name is a full id, for example en-US-Harper:MAI-Voice-2.1, and styles are applied through SSML on the Azure Speech SDK or REST.
A Sume alternative for scripted lines
A voicemail or IVR prompt is a short scripted line, so character count sets the cost. At Sume's $0.0475 per 1,000 characters, a 400-character prompt is $0.019. The Sume endpoint has no SSML or style field; you give a plain transcript and a voice.
- 400 characters x $0.0475 / 1,000 = $0.019.
- 20 prompts of that size = $0.38.
- Re-record a single line without paying for the rest.
Caveat
MAI voices are in public preview with no SLA, per the same page, so keep a fallback for anything customer-facing.
Sources
Related posts
More in Models
- Which Korean voices does MAI-Voice-2.1 have, and with what styles?
MAI-Voice-2.1 lists four ko-KR voices: Grant, Harper (role styles), Haena (14 emotion styles) and Junho (12). Pair a voice with the right caption style.
- mask_url returns 400 on Nano Banana 2 and Ideogram 4.5: your options
Only ChatGPT Image 2.5 accepts mask_url on Sume; other models answer 400 unsupported_parameter. Three ways to still change one region of a photo.
- Mercury Voice vs text to speech: what a voice agent LLM does not do
Inception's Mercury Voice writes replies for voice agents. It does not synthesize audio. Where TTS and STT jobs sit around it, with Sume rates for each.
- MiniMax H3 Max: the prompt-adherence variant on Sume
fal describes MiniMax H3 Max as tuned for prompt adherence. On Sume, minimax-h3-max runs 480p to 1080p for 5 to 15 s with frames and references.
Written by Sume