帧跃科技
Category: AI in Media / Gaming / Entertainment
An AI interactive video startup building a "trinity architecture" platform that unifies generative models, code-based generation, and physics rendering for enterprise video creation. 帧跃科技 was founded in 2025. The company is led by 杨昌鹏 (Yang Changpeng). Based in Hangzhou, Zhejiang, China. Team size: 11-50. Total funding raised: $10.0M. Latest round: Angel. Key investors include 创新工场 (Sinovation Ventures), 国谦资本 (Guoqian Capital), 零一万物 (01.AI), 璞跃中国 (Plug and Play China), 盈动资本 (Yingdong Capital), G5 Capital.
- Founded
- 2025
- Headquarters
- Hangzhou, Zhejiang, China
- Team size
- 11-50
- Total funding
- $10.0M
Value proposition
A unified "trinity architecture" that dynamically orchestrates between generative AI models, code-based generation, and physics rendering — selecting the optimal approach per task for the best balance of visual quality, logical control, and physical consistency.
Products and solutions
Leadde — AI interactive video platform (launching H2 2026) that unifies three generation paths: DiT model-based video generation, code/programmatic rule-based generation, and 3D space/physics real-time rendering. Targets enterprise and professional users for training, onboarding, tutorials, marketing, and educational videos. Supports 88 languages, 175 dialects, 300+ AI avatars, document-to-video conversion (DOC, PDF, PPT, text).
Unique value
"三位一体" (Trinity) architecture combining DiT generative models, code-based procedural generation, and 3D physics rendering in a single platform — dynamically orchestrating across all three based on task requirements.
Target customer
Global enterprise and professional users: corporate training teams, L&D departments, marketing teams, educational institutions, sales enablement teams.
Industries served
Corporate Training, Education, Marketing, Sales Enablement, IT/Cybersecurity, Healthcare, Real Estate, Financial Services, Manufacturing, Government
Technology advantage
Founding team combines rare dual expertise: CEO Yang Changpeng brings deep real-time interactive media infrastructure, rendering, and physics engine experience from Huawei Cloud; CTO Li Minglei brings multimodal LLM expertise as former Huawei Cloud multimodal chief scientist who led the original Pangu model training. The "trinity architecture" is a novel approach not seen in competitors.
How they differentiate
Unlike pure generative video models (Sora, Kling) that lack control and consistency, or pure avatar platforms (Synthesia, HeyGen) that lack generative video capabilities, 帧跃科技's trinity architecture dynamically selects between model generation, code generation, and physics rendering based on the specific task — offering both high visual quality and precise logical control.
Main competitors
Synthesia, HeyGen, Colossyan
Key partnerships
Huawei (founders are ex-Huawei Cloud executives, "华为系" spinout)
Notable customers
Fortune 500 companies (mentioned on Leadde website as "Trusted by Fortune 500 Companies" but specific names not disclosed)
Major milestones
2025-05: Company founded (registered in Hangzhou), 2026-05: Completed ~$10M+ Angel round led by Sinovation Ventures and others, 2026-H2: Planned launch of Leadde platform product
Market positioning
Early-stage AI startup positioned at the intersection of generative video AI and enterprise avatar-based video platforms, targeting the "interactive video" niche that combines generative capabilities with structured, controllable video production.
Geographic focus
Global (with presence across Singapore, Silicon Valley, and Beijing)
Patents and IP
CTO Li Minglei holds 50+ patents; team has 40+ CCF-A publications, 3 best paper awards, 3 world championship wins.
About 杨昌鹏 (Yang Changpeng)
Ex-Huawei Cloud Media Innovation Lab Founding Director & Interactive Media Lead #1; NTU & UC Berkeley joint PhD; won Franz Edelman Award (described as "Nobel Prize of resource scheduling") for media infrastructure breakthroughs.
Official website: https://leadde.ai