Video Render

local

video-render

Produce Hyperplexed-style short videos (motion graphics + AI voiceover, no face) directly from chat. Interactive flow: plan_video_scenes → user approves in chat → generate_narration → render_video → user can discard_video if not happy.

Included in every Nova Skills API key. Nothing to install — get a bearer token and this skill is available on your MCP endpoint.

Tools (4)

plan_video_scenes
PLAN a short video (Hyperplexed-style motion graphics, no face). Returns a ScenePlan JSON that the user can review and edit in chat before anything is rendered. Call this BEFORE generate_narration or render_video. Fill the schema based on the user's topic — write a script sized to the target duration (~150 words per minute), pick 3-8 scenes, and use screenshot scenes only when the user has attached (or previously uploaded) an image whose documentId you know. The user sees the plan inline and will say 'go' (or edit) before you render.
generate_narration
SYNTHESIZE narration audio from the script (default provider: msedge-tts, free; upgrades to ElevenLabs if ELEVENLABS_API_KEY is set on the server). Call this AFTER the user approves the ScenePlan. Saves an MP3 to My Documents (permanent) and returns its documentId + duration.
render_video
RENDER the final MP4. Takes an approved ScenePlan and the narration documentId from generate_narration, composites everything with Remotion server-side, and saves the video to My Documents. Returns the video documentId — the UI will play it inline. If the user dislikes the result, call discard_video with the documentId to remove it.
discard_video
Delete a video (or any Document) the user does not want to keep. Use after render_video if the user rejects the output.