Video and Image Generation: Sora, Veo, Kling, Flux — 2026-10-16

Top developments

Kling 4.0 Launches with 30-Second Native Generation

On October 15, 2026, Kuaishou’s Kling AI officially launched Kling 4.0, a major upgrade focused on high-dynamic scenes, strong camera control, and professional-grade outputs. The new model supports native 30-second video generation and 10-bit HDR output, addressing previous limitations in clip length and color depth. This release intensifies competition in the video generation market, positioning Kling as a direct rival to high-end models like Veo 3.1 and Runway Gen-4 for professional creators requiring longer, cinematic clips.

FLUX 3 Image Klein Released with Open Weights

Black Forest Labs released FLUX 3 Image Klein on October 15, 2026, a 5.6 billion parameter version of the FLUX 3 Image model with open weights. This model retains core features such as text-to-image generation, image editing, multi-reference support, and bounding box integration. The release of open-weight variants continues to democratize access to high-quality image generation tools, allowing local deployment and fine-tuning by developers outside the API ecosystem.

Midjourney Identified as "Worst Offender" in Copyright Report

A new report released on October 14, 2026, identified Midjourney as the "worst offender" among AI models for flouting Hollywood intellectual property rights. This finding follows Disney’s earlier accusation that Midjourney is a "bottomless pit of plagiarism". The report underscores the growing tension between generative AI developers and content owners, with more than 18 months having passed since initial lawsuits without significant resolution.

Appeals Court Rules Against Fair Use for AI Training Data

In a significant legal development reported on October 10, 2026, a US appeals court upheld a ruling that using copyrighted material to train AI models does not constitute "fair use". The case involved Ross Intelligence, an AI legal research company, whose training practices were deemed infringing. This precedent could have far-reaching implications for video and image generators like Runway and Midjourney, which rely heavily on large-scale datasets scraped from the internet.

Local view

Chinese media outlets are closely following the launch of Kling 4.0, with Sina Finance and 21st Century Business Herald emphasizing its appeal to professional users through enhanced controllability and HDR support. Local stakeholders note that Kling 4.0’s ability to generate 30-second clips natively addresses a key pain point for short-video creators who previously had to stitch together shorter segments. Meanwhile, user complaints on social media platforms like Sina Weibo highlight ongoing wait-time issues for free-tier users of competitors like Jimeng (ByteDance), where users report waits of up to six hours for simple generations.

Context & numbers

The Artificial Analysis leaderboard for Text-to-Video models (AA-Video-T2V v2.1) currently ranks Wan 3.0 first with an Elo rating of 1135, followed by MiniMax H3 (768p) at 1121. In the Image-to-Video category (AA-Video-I2V v1.1), MiniMax H3 (768p) leads with an Elo of 1141, followed closely by Dreamina Seedance 2.5 at 1139. These rankings reflect the competitive intensity among Chinese tech giants (Alibaba, Tencent, ByteDance) and international players. Pricing for Luma Ray 3.2 remains at $0.30 per 5-second clip at 720p and $1.20 at 1080p.

On the radar

• Runway Copyright Litigation: A class action lawsuit filed by a YouTuber against Runway AI in February 2026 is ongoing, alleging misuse of YouTube content for training; updates on settlement or discovery phases may emerge soon.
• FLUX 3 Video Early Access: Black Forest Labs has indicated that video capabilities for FLUX 3 are in gated early access, with no public pricing yet announced; industry watchers await broader availability details.