animate-avatar

Generate animated videos from static avatar images using FAL AI. Create looping work and relax animations with Kling V2.6 Pro, LTX, or Veo models.

SKILL.md

GitHub

Animate Avatar Skill

Generate animated videos from static avatar images for use in apps, websites, or social profiles.

Overview

This skill automates the process of:

  1. Generating multiple video animations from a static avatar image
  2. Optimizing videos for web delivery

Activity Philosophy

Default Generation: Create 3 videos per avatar:

  • 2 Work activities - Shows the character being productive
  • 1 Relax activity - Shows the character taking a break

This 2:1 ratio of work-to-relax creates variety and humanizes characters.

Work Activity Prompts (pick 2)

  • "Character typing on a computer keyboard"
  • "Character writing or signing documents"
  • "Character giving a presentation with gestures"
  • "Character thinking deeply, tapping chin"
  • "Character nodding while listening attentively"
  • "Character examining something closely"
  • "Character taking notes with a pen"

Relax Activity Prompts (pick 1)

  • "Character drinking coffee or tea"
  • "Character playing a small guitar"
  • "Character eating a snack happily"
  • "Character stretching and yawning"
  • "Character listening to music, bobbing head"
  • "Character waving hello cheerfully"

Models & Pricing

Model Flag Cost Notes
Kling V2.6 Pro --model kling $0.07/sec ($0.35/5s video) Default, best quality/cost
LTX-2-19B --model ltx ~$0.20/video Cheaper for small res
Veo 3.1 --model veo $0.20/sec ($1.00/5s video) Highest quality

Prerequisites

  • FAL AI API key set as FAL_KEY environment variable
  • ffmpeg installed for video optimization

Usage

bun .claude/skills/animate-avatar/scripts/animate-avatar.ts \
  --avatar "avatar-name" \
  --prompts "Character typing on keyboard,Character signing a document,Character drinking coffee"

Parameters

Parameter Required Default Description
--avatar Yes - Avatar name or path to image
--count No 3 Number of videos to generate
--prompts No Auto Comma-separated motion prompts
--model No kling Model: kling, ltx, or veo
--skip-optimize No false Skip ffmpeg optimization

Output

Videos are:

  1. Generated via FAL AI (~70s each)
  2. Optimized with ffmpeg (320x320, ~150-250KB)
  3. Saved to the output directory

Tips for Best Results

Good Prompts

  • Keep motions subtle - loops better
  • Match the character's personality
  • Work: typing, writing, presenting, thinking
  • Relax: drinking, stretching, waving

Avoid

  • Large movements that break the loop
  • "Walking", "running", "jumping" - too dynamic
  • Complex multi-step actions

Video Specifications

Property Value
Resolution 320x320
Format MP4 (H.264)
Duration 5 seconds
Target Size ~150-250 KB