Voice Assistant UX vs Gemini Live: Russian TTS, Vision Agents & CIS Market Opportunity

[zad111ak-ai]

Hello Qwen Team,
I'm an active user from the CIS region, and I want to share detailed product feedback comparing Qwen's voice assistant with Google Gemini Live. I've identified critical UX gaps and a massive strategic market opportunity.

🎯 Technical & UX Observations

1. Russian Language TTS Issues (Critical)
   The biggest blocker for Russian-speaking users is incorrect stress placement (ударения). It appears Qwen3-TTS was trained without explicit stress marks, causing almost every word to be pronounced incorrectly. This makes the voice sound robotic. I know this is a known architectural issue (e.g., Qwen3-TTS Discussion Russian language TTS issues #185), but it heavily degrades the mobile app experience.

2. Latency & Conversational Flow
   Compared to Gemini Live, Qwen's duplex communication feels less fluid. It lacks the natural pauses, "breathing", and conversational latency matching that makes a life-assistant feel human.

3. Multimodal Vision (Screen/Camera) & Ecosystem Integration
   Gemini Live and Project Astra offer a game-changing feature: real-time screen sharing and camera vision. The assistant can "see" what the user is doing on their device or looking at in the real world, providing real-time audio guidance, narrating the UI, or interacting with external objects. While Qwen has promising research on UI Agents, this seamless vision-to-voice loop is currently missing in the consumer mobile experience.
   Furthermore, Gemini relies on deep integration with Google services. Since Google's ecosystem is restricted in the CIS region, a superficial replacement isn't enough. Qwen has a unique opportunity to build an open API/plugin framework, allowing users to deeply integrate the voice agent with local alternatives (Yandex, regional banking apps, local maps, e-commerce).

💡 Strategic Market Opportunity: The CIS/Russia Gap
There is a massive, empty market for a native "Life Assistant" in Russia and the CIS:
Google Gemini and Apple Intelligence are restricted or unavailable.

Popular Android OEMs (Xiaomi, Honor, Tecno, Infinix) dominate the regional smartphone market.

If Alibaba partners with these OEMs to integrate Qwen deeply at the OS level (similar to the Honor MagicOS partnership) and solves the Russian TTS stress issues, Qwen could instantly become the de facto standard voice assistant for millions of devices in the region.

📊 Suggested Priorities

Fix Russian TTS stress marks to unlock the CIS market.
2.
Implement Multimodal Vision (Screen/Camera sharing) in the voice mode to match Project Astra's real-time assistance capabilities.
3.
Build an open plugin architecture for local CIS services to replace the unavailable Google ecosystem.
4.
Optimize voice mode latency to match Gemini Live's conversational feel.
5.
Explore OEM partnerships in the CIS region to ship Qwen as the native OS-level assistant.
Qwen's text reasoning and open-source strategy are excellent. By closing these multimodal and localization gaps, you could completely dominate the underserved CIS market. Thank you for your hard work!