Thursday was a day crowded with new moves in AI video and image generation. Platforms pushed boundaries on narrative length, editing ease, and TV integration, while others quietly shifted user bases and funding.
Novi AI’s Long Video Agent
Novi AI’s Long Video Agent stands out. It’s not just another clip generator. This agent is built to take a story idea or screenplay and output a full narrative video, up to five minutes, all in one go. The move addresses a familiar gap—longer-form video generation without breaking the process into separate steps. From script to voiceover to multi-scene visuals, Novi AI bundles it fluidly, which could change the way creators approach AI video production. For the portfolio: a true pipeline consolidation.
Google TV Integrates AI Creation Tools
Meanwhile, Google TV hijacked the day’s spotlight by embedding AI tools for image and video creation directly into its interface. Nano Banana turns voice prompts into images, letting users remix pictures live on their TVs. Veo tackles video, animating pictures or generating new clips from text. “Make my grandfather moonwalk in space” is suddenly a home command. The launch is limited to Gemini-powered TCL Google TVs in the U.S., but it signals a shift toward everyday devices morphing into creative hubs.
Buzzy Simplifies Video Editing
If editing still felt like a slog, Buzzy peeled that friction away. Their so-called “AI Video Photoshop” lets users chat their way through video edits at the pixel level. Remove a background object, correct eye contact, adjust lighting—no timeline mastery needed. That kind of chat-based control could make video tweaks as effortless as text editing. A quiet revolution in hands-off refinement.
User Migration After OpenAI's Sora Shutdown
The fallout from OpenAI’s Sora shutdown also rippled through the landscape. With Sora’s web and app versions shuttered, and the API slated for September closure, users migrated in numbers to Kling AI and Runway ML. Kling, especially, saw its active users tick up 4%, cementing its growing global presence. The market doesn’t pause when a player exits—it just redistributes.
Image Generation Model Shifts
On the image side, OpenAI is preparing to retire DALL-E 2 and 3 by mid-May, making room for GPT Image 1.5 and the forthcoming GPT Image 2. The handover reflects shifting usage as new models like FLUX and Google’s Imagen 3 dominate. Meanwhile, ComfyUI grabbed $30 million to scale out its open-source creative platform, which already supports over 4 million users across image, video, 3D, and audio generation using its node-based approach. That’s infrastructure expansion with an eye on collaboration and multi-modal creative workflows.
The Evolving Landscape of Visual AI
The day ended with a subtle but clear message: visual AI is not just about sharper pixels or deeper models anymore. It’s about blending creative stages into fewer steps, making tools more conversational, and embedding them where creation happens naturally. From the living room couch to the editing chair, the cost of making moves closer to zero.
For the archive: the quiet churn beneath the flashy launches is just as telling as the headlines.
