Microsoft MAI-Voice-2
B Tier · 7.3/10
Microsoft's in-house expressive TTS line. **MAI-Voice-2.1 and MAI-Voice-2.1-Flash launched 2026-10-01: 23 languages and 26 locales with one voice that keeps the same speaker and a native accent across languages, voice cloning from a few seconds of reference audio with consent guardrails, priced at $22 per 1M characters for 2.1 and $15 per 1M characters for Flash, which generates 45 seconds of audio at 150 ms end-to-end latency and is pitched at voice agents alongside MAI-Transcribe-2-Streaming.** These supersede **MAI-Voice-2 (Build, 2026-06-02)**: 15 languages, emotion tags, zero-shot cloning from a 5-60 s clip, preferred over MAI-Voice-1 72% of the time. Microsoft also shipped a MAI-Voice-2-Flash and MAI-Image-2.5-Pro in July that this page had not recorded -- see the Oct 2 note.
Score Breakdown
The Good and the Bad
What we like
- +Speed is the real headline -- 60 seconds of audio generated in about 1 second on a single GPU. That is a different class from ElevenLabs or Voxtral for high-volume workflows where throughput beats the last ~5% of expressiveness
- +First-party Azure Foundry integration means Microsoft customers get a TTS option that doesn't involve an OpenAI dependency. For enterprises managing AI vendor concentration, this is a real unlock
- +Already in production at scale -- powers Copilot, Bing voice, PowerPoint narration, and Azure Speech as of launch. Not a research preview that might never ship
- +Custom voice cloning from a few seconds of input is competitive with ElevenLabs, inside an Azure-native security and compliance envelope that enterprise buyers actually need
What could be better
- −Not available as a consumer subscription. API-only pay-as-you-go on Foundry means you need an Azure account and engineering work to use it -- no claude.ai-style website for casual use
- −MAI Playground is US-only at public-preview launch -- international users get pushed straight to the API
- −MAI-Voice-2 narrowed the expressiveness gap with emotion tags and 15-language support, but ElevenLabs v3 still has the deeper preset library, finer style controls, and a polished consumer UI
- −Voice cloning raises the same policy concerns as ElevenLabs -- Microsoft has enterprise guardrails but you should still be careful about consent and deepfake risk
Pricing
MAI-Voice-2.1 (launched 2026-10-01)
- ✓23 languages and 26 locales; a single voice speaks every language with a native accent (English, then Mandarin, then German, same speaker)
- ✓Voice cloning across all supported languages from a few seconds of reference audio, with built-in consent guardrails
- ✓Microsoft's strongest multilingual TTS to date per the launch post
- ✓Available through OpenRouter at launch (other channels listed in the post)
MAI-Voice-2.1-Flash (launched 2026-10-01)
- ✓Same languages and cross-language speakers as 2.1, tuned for high-volume, latency-sensitive workloads
- ✓Generates 45 seconds of audio with 150 ms end-to-end latency
- ✓Microsoft claims 55% faster inference and about 60% cheaper than comparable models
- ✓Positioned as the speech half of a voice agent next to MAI-Transcribe-2-Streaming ($0.54/hr)
MAI-Voice-2 (Azure Foundry, launched 2026-06-02)
- ✓15 languages with code-switching (Hindi-English, Spanish-English)
- ✓Granular emotion control via tags (sad, whispered, excited, etc.)
- ✓Zero-shot voice prompting from a 5-60s reference clip
- ✓Preferred over MAI-Voice-1 72% of the time; speaker similarity rated 'indistinguishable' from real recordings
- ✓Integrated into VS Code + Dynamics 365 Contact Center
MAI-Voice-2-Flash (coming soon)
- ✓Efficient, lower-cost variant of MAI-Voice-2
- ✓Announced 2026-06-02, not yet available
MAI-Voice-1 (original, 2026-04-02)
- ✓English-only expressive TTS
- ✓~60s of audio generated in ~1s on a single GPU
- ✓Reference price point for the generation pending 2.0 disclosure
MAI Playground (Free preview)
- ✓US-only web playground for testing
- ✓Rate-limited preview access
- ✓No commercial use -- evaluation only
Bundled (Copilot / Bing / PowerPoint / Azure Speech)
- ✓Existing Microsoft 365 Copilot subscriptions use MAI-Voice-1 under the hood
- ✓No separate configuration or pricing required for existing Microsoft customers
Known Issues
- MAI-VOICE-2.1 AND 2.1-FLASH -- 23 LANGUAGES ON ONE VOICE, $22 AND $15 PER 1M CHARACTERS, AND THE FIRST PUBLISHED PRICES FOR THIS LINE (2026-10-01, vendor-primary): alongside MAI-Transcribe-2-Streaming, Microsoft AI announced 'two new voice models: MAI-Voice-2.1 and our blazing-fast variant, MAI-Voice-2.1-Flash'. **MAI-Voice-2.1** is 'our strongest multilingual text-to-speech model yet', expanded to **23 languages and 26 locales** 'while enabling one single voice to use all languages with a truly native accent' -- the same speaker stays recognisable when switching from English to Mandarin to German -- and is **'priced at $22 per 1M characters'**. **MAI-Voice-2.1-Flash** keeps the languages and cross-language speakers but is 'leveled up for high-volume, latency-sensitive workloads': **45 s of audio at 150 ms end-to-end latency**, '55% faster model inference and ~60% cheaper than comparable models', **'best-in-class pricing of $15 per 1M characters'** (comparable-model claims are Microsoft's, unnamed). Both support **voice cloning across all supported languages 'using just a few seconds of reference audio'** with 'built-in consent guardrails'. Distribution: 'You can get to work with MAI-Voice-2.1 and MAI-Voice-2.1-Flash through OpenRouter, and all three models through' a list our extract truncated -- Azure Foundry is likely from precedent but not verified here. **Price context for this page:** this is the first dollar figure Microsoft has published for MAI-Voice (the June MAI-Voice-2 entry below says 'Not disclosed'); ElevenLabs' Eleven v4 is sold by plan credits, Gemini 3.8 Flash TTS is $0.50/$9.00 per 1M tokens (not characters), and xAI's grok-tts is $15 per 1M characters -- so Flash matches xAI's rate and 2.1 sits above it. **Staleness note:** Microsoft's blog index also lists 'Introducing MAI-Image-2.5-Pro and MAI-Voice-2-Flash' (July 2026), which this page never recorded; 2.1-Flash now supersedes it, so the July post was not separately opened this sweep.Source: Microsoft AI (microsoft.ai/news/our-first-streaming-transcription-model/, JSON-LD datePublished 2026-10-01T16:00Z) + microsoft.ai/blog/ index via Firecrawl (lists the July MAI-Voice-2-Flash post) -- fetched 2026-10-02 · 2026-10-01
- VERSION BUMP (2026-06-02, Microsoft Build): MAI-Voice-2 launched in the 'seven new MAI models' wave. Vendor-published changes vs 1.0: expanded from English-only to 15 languages (incl. code-switching for Hindi-English and Spanish-English); granular emotion control via inline tags (sad, whispered, excited, etc.); zero-shot voice prompting from a 5-60s reference clip; improved speaker consistency across long-form content. Preference testing: listeners preferred MAI-Voice-2 over MAI-Voice-1 72% of the time, and in speaker-similarity evaluation its output was rated 'indistinguishable from recordings of the same voice' -- across 11 languages 45.5% of listeners preferred the synthetic speech vs 44% for human recordings. Now on Azure Foundry and being integrated into VS Code and the Dynamics 365 Contact Center. A lower-cost MAI-Voice-2-Flash is coming soon. Per-character pricing not disclosed at launch.Source: Microsoft AI (microsoft.ai/news/mai-voice-2expressive-speech-in-10-languages/, microsoft.ai/news/building-a-hillclimbing-machine-launching-seven-new-mai-models/) · 2026-06-02
- Public preview in US only for MAI Playground. International Foundry API access works but you need an Azure subscription to testSource: Microsoft AI launch post, Tech Community blog · 2026-04
- Prior-sweep research incorrectly attributed a FLEURS WER #1 claim to MAI-Voice-1. That claim applies to MAI-Transcribe-1 (transcription), not Voice-1 (TTS). Voice-1's headline is speed, not WERSource: Microsoft model card corrections · 2026-04
Best for
Microsoft shops already on Azure who want a TTS option without an OpenAI dependency. Also good for any high-volume TTS workflow (audiobook batch generation, voicemail systems, IVR, bulk narration) where the 60x-faster-than-realtime speed beats ElevenLabs v3's slightly more expressive output.
Not for
Consumer creators who want a polished web UI with presets and style controls -- use ElevenLabs. Also not ideal if top-quartile emotional expressiveness (laughter, sighs, dramatic reading) is your requirement -- v3 still wins there.
Our Verdict
MAI-Voice-2 (2026-06-02) turns Microsoft's speed-first TTS into a genuinely well-rounded one. The April MAI-Voice-1 traded expressiveness for throughput; 2.0 keeps the speed story but adds 15 languages, inline emotion tags, and 5-60s zero-shot voice cloning -- and the preference data is striking: listeners picked it over MAI-Voice-1 72% of the time, and across 11 languages slightly preferred its synthetic speech (45.5%) to actual human recordings (44%). With integration into VS Code and Dynamics 365 Contact Center, it is now Microsoft's default voice layer. ElevenLabs v3 still wins on preset depth and a polished consumer UI, but for Azure shops the case for a third-party TTS line item keeps shrinking. The open question is per-character pricing, which Microsoft did not disclose at launch -- and the cheaper MAI-Voice-2-Flash is still to come.
Sources
- Microsoft AI: Our first streaming transcription model debuts at no. 1 on Artificial Analysis (2026-10-01) -- MAI-Voice-2.1 at $22/1M characters, MAI-Voice-2.1-Flash at $15/1M characters, cloning with consent guardrails (accessed 2026-10-02)
- Microsoft AI: MAI-Voice-2 -- expressive speech in 15 languages (2026-06-02) (accessed 2026-06-02)
- Microsoft AI: Launching seven new MAI models (2026-06-02) (accessed 2026-06-02)
- Microsoft AI: 3 new MAI models in Foundry (accessed 2026-04-17)
- Microsoft Community Hub: MAI models in Foundry (accessed 2026-04-17)
- MAI-Voice-1 Foundry model card (accessed 2026-04-17)
Explore more Microsoft MAI-Voice-2 rankings
Deeper leaderboards, benchmarks, task-specific tier lists, and status/pricing pages for Microsoft MAI-Voice-2.
The Tier List Tuesday
Weekly newsletter: tier movers, new entrants, and the VS of the week. Built from our daily AI-tool sweeps. No spam, unsubscribe anytime.
Alternatives to Microsoft MAI-Voice-2
ElevenLabs
**Instant Soundtracks (2026-10-15) scores any uploaded video with original music and timed sound effects in ElevenCreative, ElevenMusic and the API, with effects on separate layers and the option to use the clip's existing audio as a style reference.** **ElevenAgents gained two new tools in the week of 2026-10-06: Architect (alpha, 10/06), a built-in expert that diagnoses agent failures from transcripts and drafts tested fixes, and synthetic voice detection (10/09), which flags AI callers in the first seconds of a call -- enterprise-only for now. ElevenCreative added Brand Kit (10/09).** **Two days after the v4 launch, ElevenLabs closed a $300M employee tender at a $22B valuation (2026-09-30), double its February Series D, with enterprise now 55% of revenue -- and the pricing page is running v4 launch promos (extra v4 credits on Creator plans and above, see the Oct 2 note).** **Eleven v4 and Eleven v4 Turbo launched 2026-09-28 -- a fourth-generation voice model on an entirely new architecture**, ranked #1 on Artificial Analysis' Provider Voice Arena and preferred by ~75% of listeners in blind head-to-head tests against Cartesia Sonic 3.6, Inworld TTS-2 and Google's Gemini 3.8 Flash and Flash-Lite TTS; v4 Turbo answers with a ~100ms median inference latency (~150ms to first speech over WebSocket) and is co-optimized with ElevenAgents. Both cover 90+ languages, follow inline audio tags like [laughs] or [phone buzzing] more accurately, clone voices with better speaker similarity, and are live now in ElevenAgents, ElevenCreative and the API (Text to Dialogue). Eleven v3 becomes the previous generation. The 2026-09-10 Universal Music Group licensing deal and CLI v1 (8/24) still stand.
Murf AI
Text-to-speech that actually sounds like a real person read your script -- not a robot trying its best
Descript
Edit audio and video by editing text -- the 'Google Docs of media editing' actually lives up to the hype
Speechify
Text-to-speech reader that turns articles, docs, and PDFs into natural-sounding audio
Grok Speech (STT + TTS APIs)
**Grok Voice 3.0, xAI's new speech-to-speech model, shipped 2026-10-14 at the same $0.08 per minute as 2.0, which xAI's rate card now flags as on its way out; grok-voice-latest switched over automatically, and xAI cites a #1 score of 86.9 on Artificial Analysis' Speech-to-Speech Index.** xAI's standalone voice APIs. **Grok Voice Transcribe 2.0 shipped 2026-09-18** -- xAI calls it twice as accurate as 1.0 at the same $0.10/hr batch and $0.20/hr streaming, ranks it first for accuracy among 32 streaming models on Artificial Analysis, and has already made it the default STT model; **1.0 is being retired in the coming weeks, so pin grok-voice-transcribe-1.0 if you need it**. **PRICE CORRECTION: text-to-speech is $15.00 per 1M characters on xAI's live rate card (verified 2026-09-21), not the $4.20 this page carried since April.** Speech-to-speech agent $0.08/min, 26 flagship voices, ~1-minute voice cloning, no-code Voice Agent Builder
GPT-Live (ChatGPT Voice)
**GPT-Live-1.1 and a smaller GPT-Live-1.1 Quip began rolling out in ChatGPT Voice on 2026-10-15, with more natural accents, shorter spoken cues while it searches, and new voices for French, German, Hindi, Japanese, Korean and Mexican Spanish plus an English voice called Bubbie.** OpenAI's full-duplex voice models (launched in ChatGPT 2026-07-08) -- listens and speaks at the same time, backchannels naturally, and delegates hard questions to a backend model while keeping the conversation going. **GPT-Live-1 reached the API on 2026-09-10 at $0.05 per minute** for the voice layer, with telephony support, keyword biasing, delegation to any backend model (GPT-6 Astra, Luna, or a third-party model) and a wider voice set. In ChatGPT: GPT-Live-1 for paid tiers, GPT-Live-1 mini for Free. **Five days after the API launch, Google priced Gemini 3.8 Live at $0.005/min audio in and $0.018/min audio out (2026-09-15)** -- a fraction of GPT-Live-1's $0.05/min voice layer, though each stack bills backend or text tokens on top
Cohere Transcribe
Cohere's first audio model -- launched 2026-03-26 under Apache 2.0, 2B parameters, #1 on Hugging Face Open ASR Leaderboard (5.42 avg WER), 14 enterprise-critical languages. Free API with rate limits; Model Vault for production