CosyVoice Studio is Alibaba's one-stop voice productivity platform for people who record, transcribe and publish audio all day: consultants, journalists, podcasters and phone-support teams. It launched on 7 August 2026 and stacks three apps on top of Alibaba's Qwen-Audio speech models, which currently sit first on the Artificial Analysis leaderboards for speech recognition, real-time interaction and text-to-speech.
Core Features
- CosyFlow dictation strips filler words and spoken self-corrections, then reshapes rambling speech into an email, a status report or meeting minutes.
- Up to six hours of continuous recording per upload, with speakers separated by voiceprint plus automatic chapters and a summary.
- CosyAgent builds a real-time voice agent from a plain-language brief and connects it to internal documents, databases, MCP servers or your own API.
- CosyCreative turns a document or a URL into a two-host podcast or a multi-role audiobook using 1,000+ voices, with cloning, partial regeneration and multi-track editing.
- Spoken figures land as digits rather than words, and one tap translates a finished transcript.
Best For
- Meeting, interview and lecture notes where speakers have to be labelled automatically.
- People who think out loud faster than they type and want clean prose, not a raw transcript.
- Support and sales teams prototyping a phone agent before committing engineers to a custom stack.
- Creators recycling newsletters and reports into podcast episodes.
Pricing
Free at $0 for individual use on iOS, Android, macOS and Windows, with no usage cap during the current promotional window. CosyAgent and CosyCreative stay whitelist-only for business accounts, and API calls are billed separately on Alibaba Cloud.
Our Take
Best for one-person media operations and note-heavy consulting work. The trade-off: the two most interesting modules are still invite-gated, and narrower rivals such as ElevenLabs on cloning or Granola on meetings go deeper at a single job. If you want the model rather than the app, read our listing for Qwen-Audio-3.0-Realtime on aifreetool.site.
