SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

Developer Builds Stateful AI Video-Editing Tool Using Google Gemini and MCP

0
·1 views

A developer has released omni-skill-agy, an open-source project that wraps Google's Gemini Omni Flash video model in a FastMCP server and packages it as a skill for the Antigravity CLI. Unlike most video-generation tools, which are stateless and forget context after each prompt, the system uses Google's Interactions API to retain visual context across multiple editing turns. Users can generate a video from text, animate a still image, interpolate between keyframes, or restyle existing footage — all from the terminal. Each generation call returns an interaction ID, allowing iterative refinements such as changing lighting or style without re-describing the entire scene. The project also supports uploading finished videos directly to YouTube, and is designed to handle practical constraints like slow synchronous generation and large file sizes via the Gemini File API.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Developer Builds In-Browser AI Background Remover Without Any Server Uploads

A developer built a fully client-side background-removal tool for their website, avoiding any server uploads by running AI models locally via WebAssembly. An initial library was dropped after its AGPL license posed legal risks for closed-source commercial use, highlighting the importance of checking AI library licenses before integration. The final solution used Hugging Face's Transformers.js with an Apache-2.0-licensed CNN model, chosen over a transformer-based model that caused memory crashes during WASM inference. A residual color halo around cutouts was fixed through alpha-curve post-processing, pushing most pixels to either fully transparent or fully opaque while preserving soft edges. The project also encountered a browser-freezing bug during computation, underscoring the real-world challenges of running machine learning workloads entirely in the browser.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Builds Stateful AI Image-Editing Skill for Google Antigravity Using Gemini API

A developer has released nb2lite-skill-agy, an open-source project that wraps Google's Gemini image model inside a FastMCP server and integrates it as a skill for the Antigravity CLI tool. The project leverages Google's Interactions API, which preserves visual context between turns, allowing users to iteratively edit an image with short follow-up prompts instead of re-describing the full scene each time. The underlying model, gemini-3.1-flash-lite-image, supports sub-two-second image generation and text rendering in over 25 languages. The integration is built on the Model Context Protocol, an open standard that lets any compatible AI assistant discover and use the tool without custom per-client code. The repository ships both an MCP server exposing four tools and a skill definition that instructs Antigravity when and how to invoke them during a coding session.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

AI Agents Shift From Conversation to Real-World Task Execution

OpenAI has announced a new phase in the development of AI agents, marking a significant evolution beyond simple chat-based interactions. The focus is now on operational capability, enabling AI systems to perform tasks and take actions in the real world rather than just responding to queries. This transition represents a broader industry move toward agentic AI that can autonomously execute multi-step workflows. The announcement signals OpenAI's intent to position its models as active operators within digital environments, not merely conversational tools.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer releases topolines, a zero-dependency React library for animated topographic backgrounds

A developer has published topolines, an open-source React component that renders animated topographic contour line backgrounds using WebGL shaders. The library generates smooth, drifting contour animations entirely on the GPU, keeping CPU usage negligible at any screen resolution. It requires no external dependencies and supports customization through props for speed, scale, line width, colors, and an interactive cursor effect. The component is SSR-safe, pauses when offscreen or the tab is hidden, and respects the prefers-reduced-motion accessibility setting. Released as v0.1, the project includes a live playground where users can adjust settings and export the corresponding code.

0 comments Read more at DEV Community