What it is
Qwen3‑TTS is an Apache‑2.0 licensed open-source text-to-speech model series with voice design and voice cloning, also available via Alibaba Cloud (DashScope/Model Studio) APIs.
Gabriel’s notes
Quick take: Qwen3‑TTS is the “wow, this is really good” kind of open-source TTS—*including* voice cloning. That’s great for legitimate products and research, and also… a reminder that reality is now optional.
Qwen3‑TTS is an open-source series of text-to-speech (TTS) models from the Qwen team (Alibaba Cloud) that supports voice design, voice cloning, streaming generation, and instruction-driven control (tone, rhythm, emotion/prosody). The repository describes support for 10 major languages (Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian) and includes multiple released model variants (e.g., 0.6B and 1.7B). It’s released under the Apache-2.0 license. It can be run locally (via the project code / Python package) or called via Alibaba Cloud’s DashScope/Model Studio API.
I saved this under Video & audio because… speech is now “generate-on-demand,” and voice identity is now a feature flag—so we need tooling and a spine.
Good fit if you want to:
- Prototype multilingual TTS with a modern, controllable model family (and not just a single “robot voice”).
- Do legitimate voice cloning for a consenting speaker (voice bots, accessibility, dubbing, localization, etc.).
- Run TTS locally for privacy, latency, or cost control (assuming you can afford the compute).
- Offer real-time/streaming speech generation in an app workflow.
- Experiment with voice design (creating a voice from a text description) and instruction-based style control.
Pricing snapshot (auto-enriched):
Open-source/local: The model code/weights are available under Apache-2.0, so the “license price” is effectively $0 (your GPU bill is another story).
API (Alibaba Cloud Model Studio): The Model Studio pricing page lists Qwen3‑TTS models billed per input text character (output not billed for those Qwen3‑TTS variants). For example, qwen3-tts-flash is listed at $0.10 per 10K characters in International deployment mode, with a 10,000-character free quota (validity details are described on that pricing page). Voice enrollment (“voice cloning” as a separate billing item) is listed at $0.01 per voice (with a free quota noted for International mode), and voice design is listed at $0.20 per voice (also with a free quota noted for International mode).
Work-use / compliance snapshot (auto-enriched):
Vendor policy details: Unknown / not confirmed (from the sources reviewed here) for a single, authoritative “voice cloning acceptable-use” policy specific to Qwen3‑TTS. Assume your organization still needs a consent + disclosure posture.
Practical reality: If you can clone a voice from a short reference sample, you can also impersonate people. So at work: get written consent from the voice owner, keep an audit trail of reference audio provenance, restrict who can generate “named voices,” and add a disclosure layer in any user-facing product.
Alternatives (auto-enriched):
- Qwen‑TTS‑Realtime (Model Studio): If you want a vendor-hosted, real-time experience, Model Studio lists a “realtime” TTS option billed per input/output tokens—more “API product” than “run it yourself.”
- Alibaba Cloud Intelligent Speech Interaction (Speech synthesis): If you want a more traditional cloud speech synthesis SKU, Alibaba’s ISI product pricing describes “speech synthesis” billed by call volume (tiered) and includes free-trial rules—different packaging than the Qwen3‑TTS models.
Before you adopt it:
- Write the consent rule first. Don’t let “we can” outrun “we should.” Make consent explicit, stored, and reviewable.
- Decide local vs API early. Local runs = more control; API runs = faster time-to-demo but you inherit vendor billing + network + account governance.
- Instrument for abuse. Log generations (who/when/what voice preset), rate-limit, and add internal review for high-risk use cases.
Sources:
https://github.com/QwenLM/Qwen3-TTShttps://www.alibabacloud.com/help/doc-detail/2881635.htmlhttps://www.alibabacloud.com/help/en/model-studio/model-pricinghttps://www.alibabacloud.com/help/en/isi/product-overview/pricing