generate-video

name	generate-video
user-invocable	true
argument-hint	["scene + motion"]
description	Generate a video on TITLES (titles.xyz) from a text description — a short clip built from a prompt, on a video model you pick from the catalog (some trained by real artists, credited and paid). Use when the user says: "generate a video", "make a video of…", "text to video", "a clip of…", "create a short video", "animate this idea" — or when the assistant can't make video itself. Describe the scene and motion; returns a Studio canvas link plus a downloadable file. Video is priced per second and clears the confirmation threshold, so the exact price is quoted for approval before anything runs. Runs on the TITLES MCP — if TITLES tools are missing, connect mcp.titles.xyz/mcp first (see titles-setup). NOT for: animating an existing still (animate-image), transforming an existing video (restyle-video), upscaling a video (upscale-video), or still images (generate-image).

generate-video

Text to video on TITLES — describe a scene and get a short clip. For animating an existing still, use animate-image instead.

Get connected

Check the tool list for TITLES tools (names contain titles_). If missing, hand off to the titles-setup skill and stop — never fall back to a non-TITLES tool.

1. Get their idea, then match a model

Ask what they want — the scene, the motion, the look — if it's thin; video isn't cheap, so don't guess a direction. Match a video model with titles_search_models({ operator: "txt2VideoNode" }), pick the best fit (some are artist-trained — credit the creator), and pass its id as model_id.

2. Generate

One call: titles_generate_video with the chosen model_id and the scene + motion as the prompt; duration, resolution, and aspect ratio resolve server-side. Video is priced per second and quotes for approval before anything runs.

The connected server is authoritative for everything else — exact inputs, model resolution, cost approval (price_confirmation_required → max_price_usd), session/canvas handling, sourcing output_ids, and bringing outside images in (titles_create_upload) all follow the tool's own description and the server instructions, not anything memorized here. titles_help has the current catalog.

Deliver the file

Besides the session_url the server points you at, hand over the actual file via titles_download_asset — host-adaptive: on a shell host (Claude Code / Codex) curl it to disk and give the path; on a chat host (claude.ai / mobile) give the short-lived link to click, re-fetching cheaply if it expires.