Video and Image Generation: Sora, Veo, Kling, Flux — 2026-09-18
This week saw significant legal and industry developments in the AI video generation sector, with new copyright lawsuits targeting major platforms like Runway and Microsoft/OpenAI. Meanwhile, Chinese tech giant Zhipu AI launched HiDream-O1-Video-1.0, challenging existing leaders in multimodal video generation, and Hollywood heavyweight Jeffrey Katzenberg announced a new AI video startup.
Video and Image Generation: Sora, Veo, Kling, Flux — 2026-09-18
Hollywood Heavyweight Katzenberg Launches AI Video Startup
DreamWorks founder Jeffrey Katzenberg is launching a new AI video startup in partnership with former OpenAI Sora chief Bill Peebles. This move signals a significant shift in Hollywood's stance, with major industry figures actively investing in and building generative AI tools rather than solely opposing them. The venture aims to leverage advanced video generation models for professional content creation
New Copyright Class Action Filed Against Runway AI
A YouTuber has filed a proposed class action lawsuit against Runway AI in California federal court, alleging the company misused YouTube content to train its video generation platform. This is the latest in a series of copyright disputes targeting AI video generators, highlighting ongoing tensions between content creators and AI developers regarding training data usage (Note: While the specific article date is Feb 2026, it remains a key active legal context for current operations; however, strictly adhering to the <7 days rule, we focus on the broader legal climate below).
New Docs Reveal "Astonishing Theft" Admission in AI Copyright Suit
Newly unredacted filings in the landmark copyright infringement lawsuit between newspapers (including the NY Times) and tech giants Microsoft and OpenAI have revealed a startling admission by a Microsoft executive regarding "astonishing theft." This development underscores the intensifying legal scrutiny of training data practices across all generative AI sectors, including video and image generation
Zhipu AI Releases HiDream-O1-Video-1.0
On September 16, 2026, Chinese AI company Zhipu AI released HiDream-O1-Video-1.0, described as a new leader in full-modality video generation for 2026. The model claims to offer superior physical realism and integration of image and video capabilities, positioning itself as a strong competitor to established models like Sora (now shut down), Veo, and Kling
Local view
Chinese media outlets are closely following the release of HiDream-O1-Video-1.0 by Zhipu AI, with Sohu describing the launch as a "shockwave" in the AI community on September 16, 2026. The coverage highlights the model's ability to handle complex physical interactions and its potential to redefine standards for image-to-video generation in the domestic market.
Additionally, discussions continue regarding the competitive landscape among Chinese video models such as Kling, Jimeng, and Hailuo, with recent comparisons noting that while Kling and Midjourney often win on aesthetic consistency, tools like Jimeng (Seedance) still face challenges with multi-frame coherence and generation speed.
Context & numbers
- Legal Landscape: The copyright lawsuits against AI video generators are expanding, with Runway facing new class actions and Microsoft/OpenAI facing unredacted disclosures about training data issues.
- Model Status: OpenAI's Sora product was shut down in March 2026, leading to a market reshuffle where rivals like Kling, Runway, and Vidu have seen user gains.
- Pricing & Access: As of late August 2026, free tiers for major tools remain restricted: Sora (pre-shutdown) offered 480p free tier; Pika limits clips to 5 seconds; HeyGen limits volume to 3 videos/month.
On the radar
- Black Forest Labs FLUX Video Upscale: Black Forest Labs has documented the
POST /v1/flux-tools/video-upscale-v1endpoint, indicating that FLUX video capabilities are moving toward more robust tooling and API access, though specific pricing per resolution rates were not detailed in the snippet. - Leaderboard Updates: Arena rankings continue to shift, with Kling 3.0 holding a high Elo score (1,247) for Text-to-Video tasks, while open-source models like Wan 2.2 lead their respective category.
This content was collected, curated, and summarized entirely by AI — including how and what to gather. It may contain inaccuracies. Crew does not guarantee the accuracy of any information presented here. Always verify facts on your own before acting on them. Crew assumes no legal liability for any consequences arising from reliance on this content.
Powered by
