Alibaba launches a free all-in-one voice AI app, and keeps the enterprise tools invite-only

Alibaba launched CosyVoice Studio on Thursday, packaging its Qwen-Audio model into a free consumer app the company is positioning as China's first one-stop AI voice platform. The product bundles dictation, voice agents, and audio creation into a single download, then splits the most valuable halves off again.

The free consumer tier, a voice keyboard that transcribes speech, strips filler words like "um" and "uh", and rewrites the result as emails or meeting notes, is available now on iOS, Android, Mac, and Windows. The two higher-value modules, CosyAgent (a voice agent builder for customer service and telemarketing) and CosyCreative (a podcast and audiobook generator with voice cloning), sit behind a whitelist with no public opening date.

Alibaba's credibility claim rests on Qwen-Audio, which the company says ranks first on Artificial Analysis across speech recognition, real-time interaction, and text-to-speech. The ranking is vendor-cited; independent confirmation against the AA platform's current snapshot was not available. The "first in China" framing is also Alibaba's, and a broader self-hosting ecosystem has long existed around open-source TTS models.

The product architecture points to where voice AI is heading: scattered transcription and synthesis tools collapsing into end-to-end voice agents. Alibaba is shipping that shift as a downloadable consumer app, with the enterprise revenue layer still invitation-only.