PixVerse bets on real-time, interactive AI video - and a new way to make shows and games
- Alibaba-backed PixVerse launches a tool that lets users direct video as it's generated - think "cry," "dance," or "pose," happening on the fly.
- Co-founder Jaden Xie says real-time generation can enable micro-dramas users can steer and "infinite" games without fixed storylines.
- China-based teams are gaining ground with faster, lower-cost models, while OpenAI's Sora still sets the quality bar.
PixVerse just moved AI video closer to live creation. Its new model lets you control a scene while the video is still being generated. You type the action; the character reacts in seconds. Less waiting, more directing.
Co-founder Jaden Xie frames the shift in practical terms: new formats, new business models. Think interactive series where the audience nudges the plot, or games that keep generating content instead of loading the next level. The tool turns video generation from a batch process into an ongoing conversation.
Why product and engineering teams should care
Real-time control changes how you scope features. Latency, prompt design, and state management now sit alongside video quality. Content teams can prototype concepts in minutes and push them live on the same platform - PixVerse already runs a social layer with 16 million monthly active users (as of October).
Xie's target is aggressive: 200 million registered users in the first half of this year and nearly doubling headcount to around 200. The company reported about $40 million in ARR in October, and says it has the runway to focus on technology over near-term monetization.
China's speed advantage - and the trade-offs
Except for Lightricks, the top eight AI video models tracked by Artificial Analysis are from China. Many are faster and cheaper to run than premium Western models. That matters when your use case depends on throughput, not just cinematic polish.
OpenAI's Sora still sets the quality ceiling, but "is constrained by generation time and API cost," said Counterpoint's Wei Sun. Chinese teams are pushing video generation as a scalable, low-cost production rail. Last month, Shengshu claimed its TurboDiffusion framework, built with Tsinghua University researchers, can generate video 100-200x faster with minimal quality loss.
OpenAI Sora shows what top-tier fidelity looks like. The market, though, is rewarding speed and price - especially for social content, short-form formats, and high-volume pipelines.
Social distribution baked in
PixVerse integrates tools into a social-style platform. Real-time generation blurs the line between creation and distribution. Users make, share, and iterate in one place. That loop is sticky - and it can compress time-to-feedback for product experiments.
Rivals are monetizing too. Kuaishou's Kling reportedly generated nearly $100 million in the first three quarters of 2025. Competition is no longer hypothetical; it's shipping and earning.
About "AI slop" - and why quality might still win
Xie compares today's flood of mixed-quality content to the early computer graphics era. Some of it will be weak. But the tech improves, the bar rises, and the market filters. His point: utility and emotional value tend to outlast novelty.
What this means for your roadmap
- Prototype an interactive video flow. Define a latency budget (e.g., sub-2s control-to-action). Test with short scenes and clear prompts.
- Model the unit economics. Compare per-minute costs across providers and formats (social clips vs. gameplay vs. ads).
- Design a moderation pipeline. Real-time generation needs fast filters, clear policies, and audit logs.
- Stick to scenario-specific UX. Tools that solve a narrow job (training, marketing shorts, UGC templates) ship faster and monetize cleaner.
- Plan for stateful prompts. Treat character, scene, and style as first-class objects you can update mid-stream.
- Measure engagement loops. If creation and distribution live together, track edit-to-share ratio, time-to-first-like, and retention after publish.
Funding and reach
Founded in 2023, PixVerse raised more than $60 million with Alibaba leading and Antler participating. Xie says another round is close, with over half of investors from outside China. The user base skews international via web and mobile app.
Where to go next
- Benchmark fidelity vs. speed: evaluate a premium model for hero content and a faster model for volume.
- Run an internal sprint to ship one interactive pilot - a micro-drama episode or a tutorial with real-time branching.
- Set governance early: creator rights, watermarking, and usage transparency for AI-generated scenes.
If you want a curated view of tools and training for video-gen workflows, see these resources:
Your membership also unlocks:
