--- title: Lightricks and MIT Redefine AI Video Speed and Realism, Echo360 Makes Strategic AI Bet, Tencent and SoulGen Also Surge Forward summary: Lightricks and MIT redefine AI video speed as Echo360, Tencent, and SoulGen also surge forward. What it unlocks for AI marketing automation and video ad. lede: Lightricks and MIT redefine AI video speed as Echo360, Tencent, and SoulGen date: 2025-05-09 updated: 2026-04-10 authors: Team COEY image: /blog/lightricks-and-mit-redefine-ai-video-speed-and-realism-echo360-makes-strategic-ai-bet-tencent-and-soulgen-also-surge-forward.webp image_alt: Lightricks, MIT, Echo360, Tencent and SoulGen share a week of AI video news keywords: AI Video News source: "https://coey.com/resources/blog/2025/05/09/lightricks-and-mit-redefine-ai-video-speed-and-realism-echo360-makes-strategic-ai-bet-tencent-and-soulgen-also-surge-forward/" --- ## This Week in AI Video Generation: Breakneck Speed, Academic Breakthroughs, and New Industry Entrants The world of generative video AI has entered a new era of velocity and realism. In the last seven days, several major players dropped releases and made moves destined to impact the entire landscape, from open-source revolutions, to education, to NSFW innovation. Here’s the must-know roundup. ### Lightricks Drops LTXV-13B: 30x Faster Text-to-Video and Fully Open-Source Lightricks, known historically for its creative apps, shifted the axis of open-source AI video this week with the release of **LTXV-13B**. Unveiled just three days ago, LTXV-13B is an open-source 13B parameter text-to-video model the company says runs up to **30 times faster** than any comparable model in the space. The bombshell here is its accessibility: Lightricks explicitly targeted systems with standard GPUs and even lower-end hardware, opening advanced generative video to creative individuals and small businesses priced out by $10,000+ GPU requirements. Community and media coverage consistently characterize LTXV as a fast, high-quality model and a direct open-source answer to closed US and Chinese solutions. The model, weights, and code are up now for all to use, with Lightricks’ CEO calling it a direct challenge to industry titans. - Read more about LTXV-13B on VentureBeat - Official launch press release ### MIT’s CausVid: Academic Leap for Smooth, Realistic Video via Hybrid Diffusion Only three days ago, MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) along with Adobe Research announced **CausVid**, an open-source, hybrid generative video model. CausVid fuses bidirectional diffusion with fast autoregressive dynamics, allowing it to generate videos in seconds that not only look sharper but also transition more coherently from scene to scene. CSAIL claims CausVid sets a new bar for both temporal realism and fidelity on open-source video benchmarks like FVD, UCF101, and Kinetics. The source code, weights, and peer-reviewed details are all available to the public, serving as a new gold standard for video research and commercial teams alike. - MIT News: Details on CausVid - CausVid source code on GitHub ### SoulGen Reinvents NSFW AI Video With Industry-Leading Generator On May 8, 2025, SoulGen shook up the NSFW segment with its official launch of a new AI-powered adult video generator. This release provides creators and business customers with an engine for generating full-scene, realistic adult-themed video at noticeably higher fidelity and frame rate than previous tools in this controversial category. SoulGen touts major leaps in prompt control, artifact-free rendering, and workflow simplicity, with the product now live for the public and deployed by adult industry platforms. This move addresses long-standing issues in “adult content AI,” signaling an inflection point for quality and reliability in NSFW video production. - Official SoulGen NSFW AI video release ### Open-Source Scene: Alibaba Wan 2.1, Tencent’s HunyuanVideo, Academic Traction The open-source surge keeps building. While the week’s headlines belong to Lightricks and MIT, Chinese giants **Alibaba** and **Tencent** remain hot topics. Alibaba’s **Wan 2.1** is still the most advanced open-source text-to-video foundation model in China, but it became available in February, *not this week*. Likewise, Tencent’s HunyuanVideo was open-sourced in March, though this week (May 8) saw Tencent deploy new workflow guides, tools, and customization options that increase HunyuanVideo’s industrial appeal. March also saw the launch of HunyuanVideo-I2V, Tencent's image-to-video foundation model, but again these are not brand new in the past seven days, aside from incremental usability enhancements and a May 8th highlight in Chinese enterprise AI outlets. - Tencent HunyuanVideo-I2V details - Tencent’s May 8th workflow guide ### Major Enterprise News: Echo360 Acquires GoReact, Targets AI Video For Learning On May 8, Echo360 made a notable strategic move by acquiring GoReact, a top platform for AI-driven video-based skill assessment and feedback. Echo360 aims to integrate next-gen AI video assessment tools into business and higher education sectors, making AI video not just about creation but also about analytics, feedback, coaching, and compliance. This acquisition signals growing enterprise appetite for generative AI video that goes beyond content creation and into performance and skills measurement. - Echo360 GoReact acquisition: announcement ### Video AI Security: Model Impersonation, Malspam Threat Surges As the volume of open-source and downloadable video models increases, cybersecurity firm Morphisec issued an alert this week (May 7) on the rise of phishing, malware, and impersonation sites disguised as new AI video services. Users, researchers, and creative professionals are advised to verify tool authenticity before loading weights or handing over sensitive data to third-party platforms. Security vigilance is now an integral part of the AI video workflow. ### Arms Race: US Open-Source Publishers vs. China’s Giants This week, US and Chinese AI video technology moves mirrored and escalated each other. Lightricks’ new model, MIT’s academic contribution, and Echo360’s acquisition directly compete with innovations from Alibaba and Tencent, though most Chinese releases this week were incremental or ecosystem focused, not major new model drops. A new cycle of accessibility (hardware, licensing) and software modularity is emerging, with open-source and enterprise use cases vying for first place. This signals a landscape racing not only for creative supremacy but for practical impact in industry and research. ### Notable Benchmarks and Academic Signals - New models from Lightricks and MIT’s CausVid are already showing state-of-the-art results on FVD and UCF101, Kinetics open datasets, being peer-reviewed and debated by the research establishment in real-time. - Tencent’s HunyuanVideo workflow and customization updates confirm its status as a staple in Chinese academic and startup deployments.