Video Field Guide

Kuaishou's flagship AI video generation model, released January 2026. The world's first native 4K@60fps direct output, supporting smart storyboarding (up to 6 auto-switching shots) and native audio-visual sync (5 languages + multiple dialects); Artificial Analysis ELO 1241. Over 60 million creators worldwide, with Q1 revenue of $75 million.

A multimodal video generation model from ByteDance's Seed Lab, released February 2026. Uses a dual-branch diffusion transformer (DB-DiT) architecture supporting text/image/video/audio four-modality input; tops the chart with Artificial Analysis ELO 1269. Usability soared from the industry average of 20% to over 90%, hailed as an 'AI-director-grade' industrial production tool. Slid to 1225 on the Aug 2026 board

Runway's General World Model (GWM) released December 2025, codenamed WhisperThunder. Tops the text-to-video chart with Artificial Analysis ELO 1247, surpassing Veo 3 and Sora 2 Pro. Supports Director Mode 2.0 (camera motion control) and Motion Brush 3.0 (independent multi-subject motion control), generating continuous shots up to 60 seconds. Not listed on the Aug 2026 Arena board

Google's new-generation video generation model launched October 2025, with a major upgrade in January 2026. First to introduce AI-generated audio and spatial audio, supporting native 9:16 vertical output for YouTube Shorts/TikTok, up to 4K resolution. Deeply integrated across the Gemini App, YouTube, Flow and Vertex AI.

MiniMax's new-generation unified multimodal video model H3 (released 2026-07-31, replacing Hailuo 02/2.3). Joint understanding and generation across text, image, audio and video; up to 2K resolution, 15s, 24FPS output with native stereo audio; supports up to 12 reference assets and first/last-frame control. Ranked #1 worldwide in Artificial Analysis video editing; model weights open-sourced on Aug 3 (MiniMax H3 Community License).

An AI video generation model co-developed by Shengshu Tech (Tsinghua's Zhu Jun team) and Tsinghua University, released January 2026. Artificial Analysis ELO 1241, supporting 16-second 1080P video generation with audio and automatic lip sync. Focused on high consistency and 2D animation/classical Chinese style rendering; Vidu Agent turns product images into marketing videos with one click. Not listed on the Aug 2026 Arena board

A video generation LLM from PixVerse (Aishi Tech), released March 2026, ranked 2nd globally by Artificial Analysis (ELO 1239). Known for ultra-fast generation (1080P in tens of seconds), with major improvements in human realism, physics simulation and audio-visual synergy. Its exclusive realtime world model R1 was named a 'Major Scientific & Technological Achievement' at the Zhongguancun Forum. Not listed on the Aug 2026 Arena board

Alibaba Cloud's AI video generation model, bursting onto the scene in April 2026. Achieved text-to-video ELO 1301 and image-to-video 1392 on Artificial Analysis, both setting new records and leading the runner-up by over 60 points. Open-source and free, supporting text-to/image-to/multi-image-reference generation; already integrated into Alibaba Cloud Bailian and the PixVerse platform. v1.1 scores 1151 on the Aug 2026 board

Pika Labs' AI video generation tool, focused on social media short videos and creative content. In 2026 it launched the 'trend algorithm' feature, automatically analyzing trending elements on TikTok/Douyin to generate videos. Added an AI voiceover library covering 50+ languages and dialects, supporting creative effects like Pikaffects, Pikascenes and Pikaswaps.

Luma AI's AI video generation model, based on the Ray3.14 architecture. Known for fast generation and cinematic camera work, supporting simulated handheld shooting and other cinematic effects. High character consistency and natural camera motion make it ideal for rapid concept validation and capturing cinematic inspiration.

Alibaba's AI video generation tool, capable of generating a 5-second 720p video in minutes on consumer-grade GPUs. Features a pioneering 'cinematic aesthetics control system', excelling at classical Chinese semantic understanding. Deeply integrated with the Alibaba Cloud ecosystem, with API services targeting enterprise customers.

Tencent's AI video generation tool built on the Hunyuan LLM, emphasizing adaptation to Chinese-language contexts. Its API services target short-video platforms, film production, ad marketing and the gaming industry, deeply integrated with the Tencent ecosystem (WeChat Channels, Tencent Video).

Zhipu AI's AI video generation tool, excelling at structured asset generation and animation control. Supports a structured creative workflow from script to storyboard, ideal for producing knowledge, science-popularization and data-visualization video content.