
Claude writes scripts but can't render video. How an MCP connector adds real video models (Veo, Kling, Sora) inside Claude in about 2 minutes.

What it cannot do natively: produce the video itself. There is no render button.
Claude supports MCP connectors: external servers that add real tools to the chat. Krater's adds 500+ models across video, image, audio, voice and documents:
Then ask Claude for the video directly: "turn this script into a 10-second product teaser." Claude calls the Krater video tool, the clip comes back into the conversation, and your Krater credits cover it alongside images, voiceovers and everything else. See the full Krater MCP setup.
If connectors are more plumbing than you want: Krater's own video studio runs the same models behind one prompt box, and the built-in assistant routes video requests automatically: krater.ai/video.
Nothing announced. Anthropic's focus has been text, code and analysis; video generation is coming from connectors and specialized models, not from Claude itself.
Not natively. With Krater connected, yes: image-to-video is one of the tools, so a still you upload or generate can become a clip.
Not inside the chat either: OpenAI's Sora is a separate app. ChatGPT gets in-chat video the same way, through a connected tool: details here.
Whatever the underlying models produce: Krater routes to leading video models like Veo, Kling and Hailuo, so the output is the same as going to them directly, minus the setup.