VEED Fabric Video Creation

VEED Fabric lets users generate professional talking head videos with AI avatars directly from a conversation. No camera or editing skills required.

Prerequisites

This skill requires the VEED Fabric MCP server to be connected. It provides all the tools referenced below. See platform-specific setup:

• Claude
• ChatGPT
• Goose

Recommended Workflow

Follow this sequence for the best user experience:

1. Browse characters — Call list_characters to show the avatar carousel. Let the user pick.
2. Browse voices — Call list_voices with the appropriate locale. Let the user pick.
3. Ask for a script — If the user hasn't already provided one, ask what the avatar should say.
4. Confirm before generating — Call confirmfabricvideo to show cost estimate and parameters. Wait for user confirmation.
5. Generate the video — Call createfabricvideo only after confirmation.
6. Check progress — Call getgenerationstatus with the returned job ID. Videos typically take 1–2 minutes.

Always let the user make selections interactively rather than choosing defaults for them.

Tool Usage Guide

list_characters

• Returns AI avatar characters with thumbnails
• Optional gender filter: "male" or "female"
• Displays an interactive carousel widget — let users browse visually
• When the user says "show me characters", call this tool

list_voices

• Requires locale parameter (e.g., "en", "fr", "es", "en-GB")
• Optional gender filter: "Female", "Male", or "Neutral"
• Displays a voice selector widget with audio previews
• Infer the locale from the user's script language or ask if unclear
• When the user provides a script in a specific language, filter voices to that language

confirmfabricvideo

• Always call this before createfabricvideo
• Shows a confirmation dialog with: script text, selected character, selected voice, credit cost estimate
• Requires script, voiceId, and characterId
• Optional: aspectRatio ("landscape", "portrait", "square" — defaults to landscape)
• Optional: workspaceId (auto-selected if user has only one workspace)
• Do not skip this step — users should always see the cost before generating

createfabricvideo

• Generates the actual video — consumes credits
• Same parameters as confirmfabricvideo
• Only call after the user has confirmed via confirmfabricvideo
• Returns a jobId for tracking progress

getgenerationstatus

• Poll with the jobId from createfabricvideo
• Status values: queued → processing → completed or error
• When complete, displays an inline video player
• If the user asks "is my video ready?", call this tool

getcreditbalance

• Shows available vs total AI Playground credits
• Optional workspaceId parameter
• Useful when the user asks about credits or before a long video

list_workspaces

• Lists workspaces the user has access to
• No parameters required
• Only needed when the user has multiple workspaces and wants to choose one

UI Widgets

VEED Fabric provides rich interactive widgets that display automatically when tools are called:

• Characters Carousel — visual browsing of avatars (from list_characters)
• Voice Selection — voice list with audio previews (from list_voices)
• Video Confirmation — cost breakdown and confirm button (from confirmfabricvideo)
• Video Result — progress indicator and video player (from getgenerationstatus)

These widgets give users a much better experience than plain text. Prefer calling the tools that trigger them.

Common User Requests

Error Handling

• Insufficient credits — If confirmfabricvideo shows the user doesn't have enough credits, inform them and suggest a shorter script or checking their plan.
• Generation fails — If getgenerationstatus returns error, let the user know and offer to retry.
• No voices for locale — If list_voices returns empty, suggest trying a broader locale (e.g., "en" instead of "en-AU").