Video Generation

Today's Stories

1. 1

   Google launches Nano Banana 2 Lite and Gemini Omni Flash for faster AI images and video

   Google released two new AI models—Nano Banana 2 Lite for fast image generation at $0.034 per image in four seconds, and Gemini Omni Flash for video generation and editing via API at $0.10 per second of output. Both are now available to developers through the Gemini API and Google AI Studio. Nano Banana 2 Lite replaces Google's older image model and offers a lower-cost, faster option for developers who prioritize speed over quality. Gemini Omni Flash opens video generation to the API for the first time, allowing developers to combine text, images, and video in a single workflow—and Google recommends chaining both models together to quickly generate images and animate them into video.

   Gemini Omni Flash currently generates only ten-second clips, and audio references and scene extensions are not yet supported in the API. Character consistency across scene changes remains limited. Both models apply SynthID watermarks to tag AI-generated content, with verification available through the Gemini app, Gemini in Chrome, and Google Search.

2. 2

   Google rolls out Gemini Omni Flash API for enterprise video editing

   Google is launching Gemini Omni Flash, the first model in its new Omni family, through an API for developers and enterprise customers. The model enables users to edit finished video clips through conversation, rather than requiring traditional production workflows involving briefs, crews, shoots, and revisions. Making video production conversational addresses a real business pain point—enterprises avoid creating internal training videos and product explainers because of the time and cost involved in traditional production chains. A single legal text change currently forces the entire workflow to restart. By cutting those friction points, the API has the potential to unlock video creation for businesses that otherwise skip it.

   Gemini Omni Flash is now rolling out to developers and enterprise customers through the API after debuting to consumers at I/O 2026. The model's core capability is editing finished clips through conversation, not just improving text-to-video prompts.

3. 3

   Soracom Adds New Feature to "Wisora" – Handover Function for AI and Human Collaborative Response

   Soracom Adds New Feature to "Wisora" – Handover Function for AI and Human Collaborative Response

4. 4

   Google DeepMind invests $75M in A24, signaling shift from video tools to content

   Google DeepMind invested $75 million（約120億円） in indie film studio A24 this week, with the two companies partnering to develop AI tools for filmmakers. Investment in AI video and media tech startups has more than tripled over the past five years, reaching $5.6 billion（約9000億円） this year, up over 43% from 2025's annual total. The market is shifting beyond competition over video-generation technology itself. Video generation is becoming easier to build, with new models from China rivaling US and European products, and incumbents developing their own AI capabilities. This means the real value may now lie in content creation and end products, not just the software layer—a change that opens opportunities for creative studios alongside tech companies.

   VCs are beginning to fund AI-native or AI-powered studios. Promise has raised funding from backers including Andreessen Horowitz and Google's AI Futures Fund, and London-based Wonder raised $12 million（約19億円） from investors including Atomico and LocalGlobe.

5. 5

   DeepSeek R1 tops reasoning benchmarks, claims #2 open-weights model spot

   DeepSeek released R1, an open-weights reasoning model that the company says achieves top performance on several reasoning benchmarks and ranks as the #2 open-weights reasoning model. The model can handle 1M tokens of context. Open-weights reasoning models give developers and researchers direct access to AI systems that solve complex logical problems, reducing reliance on proprietary closed models. R1's performance may lower barriers for organizations building reasoning-dependent applications.

   R1 uses 27% of the FLOPs (computational operations) compared with DeepSeek-V3.2 according to the company's claims, suggesting efficiency gains that could affect deployment costs for users building with the model.

6. 6

   ByteDance launches Seedance 2.5 video AI model in early July

   ByteDance introduced five new AI models at Volcano Engine's FORCE conference, with Seedance 2.5—a video generation model—as the centerpiece. The model is set to launch in early July. Video generation capability is a competitive frontier in AI. ByteDance's move to expand its AI portfolio signals the company's ongoing effort to advance its generative AI offerings alongside other major technology players.

   Seedance 2.5 becomes available in early July. The model's specific capabilities and performance will determine its impact on the video AI landscape.