invideo

• Turning several hours of interview or vlog footage into a first cut with flubs, pauses, and filler removed, then finishing it on the same timeline
• Cutting a multicam podcast or panel where the agent switches angles to whoever is speaking
• Asking the agent to generate or find a missing b-roll shot and drop it on the timeline without leaving the project
• Building a sound bed or a stylized visual effect from a description, then adjusting the layers and controls by hand
• Producing localized, dubbed versions and vertical cut-downs of a finished ad for several platforms
• Having Claude Code or ChatGPT import footage and assemble an edit through the invideo MCP server while you watch and correct it in the browser

invideo's bet is that the agent should live inside the editor, not beside it. Because the assistant's work lands as normal clips on a timeline with real color and audio tools, you can hand off logging, first cuts, multicam, and sound beds and still make the final calls yourself, and the free manual editor makes that easy to try. The MCP server lets an agent you already use, such as Claude Code or ChatGPT, work on the same timeline you are watching, with each edit landing live under its own cursor. The catch is cost visibility. Every agent action and generation spends credits at varying model rates, so heavy generation work needs watching. Editors whose finish has to happen in Premiere, Resolve, or Avid may find Eddie AI's handoff-first approach a better fit.

About invideo

invideo has moved through three products: the template-based invideo Studio, the generator known as invideo AI, and, since 2026, invideo Editor, which it calls the agentic video editor. The editor runs in the browser with Edit, Color, and Audio rooms: a multitrack timeline with Premiere-style shortcuts, keyframes with custom Bezier easing, motion graphics, a grading suite with curves, qualifiers, LUTs, and scopes, and an audio mixer with buses and Voice, Music, and SFX routing. An assistant editing agent works inside the same project. You describe the edit in plain words and it watches the footage, then places the result as ordinary clips you can move, trim, or undo. It builds a base cut from raw footage (with a script, it flags lines that were never shot), switches multicam angles by who is speaking, finds moments by person, action, emotion, or object, matches the pacing of a reference video, and removes silences and filler. Newer jobs include Agentic Sound Design, which builds a layered, editable sound bed from a described mood, and Agentic Effects, which builds a look on your footage and exposes adjustable controls. Agent Two, invideo's conversational creation agent, runs in the same workspace to plan and generate shots, and you can set up a crew of role agents (casting, locations, scenes, music) that share project memory and work in parallel. Teams edit together in real time, with review links for frame-accurate comments, branching version control, and session replay. Outside agents can join too: the invideo MCP server lets Claude, ChatGPT, Claude Code, Codex, Cursor, VS Code, and Gemini CLI edit the timeline live under their own cursor. Projects export as MP4 or as editable projects for Premiere Pro, Final Cut Pro, or DaVinci Resolve. The main difference from Eddie AI is where the edit is finished. Eddie is an assistant that logs footage and builds cuts to hand off to Premiere Pro, DaVinci Resolve, Final Cut Pro, or Avid, and charges credits for imports and NLE exports. invideo is the editor itself: you finish on its browser timeline with manual tools that cost nothing, multiplayer editing, and generation in the same project, and NLE export is optional and free. The tradeoffs: all AI work is billed in credits priced at each model's list rate, so costs vary with the models and video length, plan credits do not roll over, and invideo can change credit costs without notice.