Skip to content
ReviewRamble

The weekly brief

AI, reviewed, in your inbox.

  • New reviews of tools, models, MCP servers and repos, with the verdict up front
  • Re-checked prices and ratings when something changes
  • New analysis pieces as they are published

Models, agents, MCP servers and the tools actually worth your time.

One email a week. Unsubscribe with the link in any brief. How we use your email

Qwen-Audio-3.1-TTS-Plus

Alibaba's closed October 2026 speech model: fourth on Artificial Analysis, $20 to $30 per million characters, four stock voices.

Reviewed by ReviewRamble editors · Updated

Share this page:XLinkedInFacebookRedditWhatsAppEmail

Our rating

4.1/5

Arena scores worthy of the top five, packaged with four stock voices and a regional catalogue that reads like a puzzle.

What is Qwen-Audio-3.1-TTS-Plus?

Qwen-Audio-3.1-TTS-Plus is the quality tier of Alibaba's Qwen-Audio-3.1 speech family. The family was announced on 23 September 2026 with upgraded recognition, speech and real-time conversation models and price cuts of about 70% for speech output; the TTS-Plus model itself appeared in Alibaba's release notes on 21 October. Alibaba positions it for audiobooks, dubbing and brand voices, and a cheaper Flash sibling for low-latency assistants.

It ships four system voices, three female and one male, and supports voice cloning, voice design from a description, plain-language delivery instructions and inline tags for laughter or breathing. Cloned voices speak Mandarin, Cantonese and about 20 Chinese dialects plus 15 other languages, including English, Japanese, Korean, French, German, Spanish and Arabic; designed voices are limited to Mandarin and English.

The model is hosted only on Alibaba Cloud Model Studio, in the Singapore and Beijing regions. No weights have been released.

How much does Qwen-Audio-3.1-TTS-Plus cost?

PlanPriceWhat you get
Singapore region$0.30 per 10,000 characters1,000,000 free characters valid 90 days
Beijing region$0.20 per 10,000 characters

$0.30 per 10,000 characters in Singapore ($30 per 1M) and $0.20 in Beijing ($20 per 1M); Chinese, Japanese and Korean characters count double. 1,000,000 free characters for 90 days in Singapore.

Is Qwen-Audio-3.1-TTS-Plus worth it?

Judged by listeners, Qwen-Audio-3.1-TTS-Plus belongs with the leaders. Artificial Analysis's main arena puts it fourth, with 1,298 Elo, behind ElevenLabs' two Eleven v4 models and StepFun's StepAudio 3 TTS. On the stricter Controlled Voice board, where every model speaks with the same cloned voices, it read second of 49 on 9 November, and first for US English accents.

That is a large step from Alibaba's open Qwen3 TTS of January, which ranks near the bottom of the same table at 928 Elo. The gap is the cost of staying closed: this model has no weights to download, and Qwen's public audio repositories stop at Qwen2-Audio.

The catch: what you can actually use is narrow. There are four system voices and no British English among them; anything else means cloning or designing your own. Price depends on region, $30 per million characters in Singapore and $20 in Beijing, and Chinese, Japanese and Korean characters count double. Artificial Analysis lists it at $19.31, a figure that matches neither published rate. Alibaba's own region lists have disagreed with one another, so confirm availability before you plan a launch around it.

Our take: for a dubbing house that will clone its own voices anyway, this is a top-five voice at a mid-table price. A team that wants to pick a ready-made voice and go will find the cupboard bare.

Where it earns its place

  • A Chinese drama studio dubbing series for Cantonese and Sichuanese audiences

    Keep each actor's voice while switching dialect

    Clone each lead once, then select the dialect by instruction rather than a separate voice. A 300,000-character script bills as 600,000 characters, about $12 in Beijing.

  • A US audiobook start-up

    Produce expressive American-English narration below ElevenLabs prices

    Use one of the four system voices or design a narrator from a description; at $30 per million characters in Singapore, a 90,000-word book of about 500,000 characters costs about $15.

Before you commit

Setup
Create a Model Studio key in the Singapore or Beijing region and call qwen-audio-3.1-tts-plus with a system, cloned or designed voice; streaming returns audio in chunks, non-streaming an audio URL valid for 24 hours.
Integrations
Alibaba Cloud Model Studio (DashScope) API with non-streaming, one-way and two-way streaming modes; PCM, WAV, MP3 or Opus up to 48 kHz.
Data and privacy
Hosted only on Alibaba Cloud; no weights or self-hosting. Retention and training terms are set by Model Studio's service terms, not stated on the model page.
Cost at scale
100 million characters a month cost about $3,000 in Singapore or $2,000 in Beijing; double that for the same length of Chinese text.
Limits
180 requests a minute per region. Connections close after 60 seconds without a new task. Four system voices; British English only on Flash.

What’s good

  • Fourth on Artificial Analysis's arena at 1,298 Elo
  • First for US English on the Controlled Voice board, 9 November
  • Cloning across about 20 Chinese dialects and 15 other languages
  • Instruction control and paralinguistic tags

What’s not

  • Only four system voices; no British English
  • Closed weights, Alibaba Cloud only
  • Region-dependent price; CJK characters count double
  • Designed voices limited to Mandarin and English

Best for

Dubbing, audiobook and brand-voice teams, especially those working in Chinese, who will clone or design their own voices and want near-top quality.

Qwen-Audio-3.1-TTS-Plus alternatives

Qwen-Audio-3.1-TTS-Plus: common questions

Is Qwen-Audio-3.1-TTS-Plus worth it?
Arena scores worthy of the top five, packaged with four stock voices and a regional catalogue that reads like a puzzle. We rate it 4.1 out of 5.
How much does Qwen-Audio-3.1-TTS-Plus cost?
$0.30 per 10,000 characters in Singapore ($30 per 1M) and $0.20 in Beijing ($20 per 1M); Chinese, Japanese and Korean characters count double. 1,000,000 free characters for 90 days in Singapore.
What are the downsides of Qwen-Audio-3.1-TTS-Plus?
Only four system voices; no British English; closed weights, Alibaba Cloud only; region-dependent price; CJK characters count double; designed voices limited to Mandarin and English.
Who is Qwen-Audio-3.1-TTS-Plus best for?
Dubbing, audiobook and brand-voice teams, especially those working in Chinese, who will clone or design their own voices and want near-top quality.
How much does Qwen-Audio-3.1-TTS-Plus cost for a team or in production?
100 million characters a month cost about $3,000 in Singapore or $2,000 in Beijing; double that for the same length of Chinese text.
Is Qwen-Audio-3.1-TTS-Plus safe for company data?
Hosted only on Alibaba Cloud; no weights or self-hosting. Retention and training terms are set by Model Studio's service terms, not stated on the model page.
What are the best Qwen-Audio-3.1-TTS-Plus alternatives?
The closest alternatives we have reviewed are StepAudio 3 TTS, Eleven v4 and v4 Turbo, Qwen3-TTS and Fun-CosyVoice3.

Answers checked Nov 21, 2026.

Make Qwen-Audio-3.1-TTS-Plus? Show this rating on your siteQwen-Audio-3.1-TTS-Plus rated 4.1 out of 5 on ReviewRamble