SketchStory — Downloadable Hand-Drawn Story Kit

Downloadable, local workflow kit — not an online video service. SketchStory is a skill you install in your own AI agent. It runs on your own computer with your own paid fal.ai key to script, illustrate and assemble hand-drawn "drawing itself" story videos. Nothing is generated or delivered by Capafy.

Before you buy: what you get

• Format: a downloadable ZIP containing one agent skill folder (sketchstory-video/: SKILL.md, Python scripts, reference docs, fonts, an example project, licence and third-party notices). No hosted service, no account on our side, no video is generated or delivered by Capafy or by ReelForge.
• You run it yourself: your own AI agent client follows SKILL.md and runs the scripts on your own computer. Every image, voice and music track is generated by fal.ai with your own API key, billed by fal.ai to you.
• Supported environment: an agent client that supports SKILL.md skills and can run shell commands. Built and tested with Claude Code; Codex, OpenClaw and similar clients should work (not separately tested). Python 3.9+, ffmpeg, Linux (tested) or macOS; on Windows use WSL. 4+ CPU cores and 4 GB+ RAM recommended, ~2 GB free disk per project, internet access to fal.ai.
• Base model: no LLM is bundled or included in the price; the kit uses whichever model your own agent client runs (built and tested with Anthropic Claude via Claude Code).
• Render time on your machine: about 6–7 minutes per minute of finished video on a 4-core machine (measured), plus a few minutes of image/voice generation.
• Required account and key: a fal.ai account with billing and an API key (FAL_KEY). Optional: an ElevenLabs key (ELEVENLABS_API_KEY) if you prefer to call ElevenLabs directly for the voice instead of through fal.ai. No other paid services are needed.

Installation

1. Download the ZIP after purchase and unzip it into your agent's skills folder (Claude Code: ~/.claude/skills/sketchstory-video/ or <project>/.claude/skills/sketchstory-video/).
2. Install Python 3.9+ and ffmpeg (macOS brew install ffmpeg; Ubuntu/Debian sudo apt install ffmpeg).
3. pip install -r requirements.txt (numpy, pillow, requests, scipy).
4. Create a key at https://fal.ai/dashboard/keys and export FAL_KEY=....
5. python3 scripts/check_env.py checks everything. Then ask your agent for a video (examples below).

Additional generation costs (paid by you to fal.ai, approximate list prices, Oct 2026)

With your own music file and standard images, a 3-minute video costs under $1.50. Provider prices can change; the kit shows an estimate before it generates anything.

What the kit does on your machine

Your agent writes the script (hook in the first 5 s, beginning–middle–payoff), generates consistent illustrated boards through fal.ai, records narration, generates or uses your music, then the kit's local code traces each board stroke by stroke, adds the colour reveal, camera moves, subtitles and chapter pills, and mixes everything with ffmpeg. Every step skips finished files, so you can fix one board or one line and re-run.

Outputs in <project>/out/: a 1920×1080 MP4 (−16 LUFS) and a small preview MP4, thumbnail.png, a 9:16 Shorts/Reels cut, chapters.txt and suggested title/description/hashtags in meta.txt. Hindi-first (Devanagari subtitles and fonts included); English and other languages work too.

How the samples were made (external services)

The sample frames, thumbnails and videos on this page were made with this workflow. Their illustrations were generated by Google Nano Banana / Nano Banana Pro models on fal.ai, the narration by ElevenLabs voices and the background music by ElevenLabs Music (via fal.ai). The stroke-by-stroke drawing animation, colour reveal, captions, thumbnails layout and final edit were produced locally by the kit's own code. Your results depend on those external services (their output quality, availability, content rules and prices) and on your prompts.

Sample videos (from the Divya Chitrakatha channel's hand-drawn Ramayan series, made with this workflow):

• EP 0 (Valmiki): https://www.youtube.com/watch?v=iNkYTaiOUyQ
• EP 8 (Lanka Dahan): https://www.youtube.com/watch?v=CZWzlMLTUiQ
• EP 9 (Ram Setu): https://www.youtube.com/watch?v=cK2lHyBTJ9Y

Example prompts (to your own agent, after installing)

• "Make a 3-minute hand-drawn Hindi video: how did Hanuman cross the ocean? Devotional style."
• "Turn this folk tale into a storybook-style video with English narration, and a 40-second Short."
• "EP 1 of a 5-part hand-drawn series on Chanakya's life, with thumbnail and description."

Styles

storybook-ink · devotional-sketch · whiteboard (explainers) · kids-crayon · noir-sketch — or describe your own.

FAQ

Is this a video service? No. It is a set of files for your own agent. You run it, on your computer, with your own fal.ai account.
Do I need to code? No, your agent runs the commands, but you need to install Python, ffmpeg and the skill, and set your API key.
Can I use my own images, music or voice recording? Yes: point a board at your own image file, or supply your own music track.
Can I monetise the videos? The videos you make are yours, subject to the AI providers' terms.
Does it clone voices? No. Use ElevenLabs library voices or voices you designed yourself.
Licence: single-buyer licence for unlimited personal and commercial use; the stroke-drawing code and hand image are from sweety-handdraw-video (MIT), noted in THIRD_PARTY_NOTICES.md.

Published by ReelForge.