I Taught My Articles to Make Their Own Reels
Six posts on the pipeline I built — voice, captions, motion, GPU, and the button that wires it together.
A few days ago, I clicked a button in my own article editor. A 55-second vertical MP4 fell out the other side.
Narrated in my cloned voice and captioned word-for-word. The background is animated by an AI model running on a 12GB GPU in the corner of my office.
The button cost me $0 per click. The pipeline behind it cost me a string of Tuesday evenings and one $5/month ElevenLabs subscription. Yes, I considered self-hosting the voice model, but more about it later.
This post is the map for the next six.
TLDR
Six-part series on the reel-generation pipeline I built into Digital Craft Workshop — one button on an article, one MP4 out, ready for IG / YouTube Shorts.
Stack: Next.js + Claude + ElevenLabs + FFmpeg + a 12GB local GPU.
Free in the series: architecture, decisions, bug arcs, illustrative code, and cost breakdowns.
Paid kit drops the week after part 6: runnable starter repo + ComfyUI workflows + Notion playbook + Loom walkthrough.
Cadence: weekly, Wednesdays. Six weeks. Subscribe below to get the rest in order.
Who this is for
Solo builders who already ship — indie devs, AI tinkerers, content creators who write more than they post. You already have an article (or a newsletter, or a script). You’re not allergic to a terminal.
You don’t want to pay $30 a month for a SaaS that takes 90% of the margin to render the same thing.
No 12GB GPU? You'll still get five of the six posts straight — Parts 1–5 run on any laptop. For Part 6, I cover Replicate, Modal, and RunPod as ~$0.30 per-scene cloud alternatives so you don't get left at the door.
I’m writing this for the version of me that started two months ago. He would have killed for someone to publish the whole arc end-to-end instead of one scattered blog post per stage.
This series is part of the broader Digital Craft Workshop project — small tools, full stack, shipped from one machine.
The series map
Each part stands alone. Each part ends on a named obstacle that the next part fixes. Read in order if you want the build-along; skip around if you came for a specific layer.
Part 1 — The whole machine in one diagram. Architecture overview. What goes in (a published article), what comes out (a 9:16 MP4), and the five stages between. The vocabulary you’ll need for everything that follows.
Part 2 — Teaching Haiku to write a screenplay. How I turn an article into a 3–5 scene JSON script with Claude Haiku 4.5. The prompt, the JSON contract, and the cache layer mean a re-render costs zero AI credits.
Part 3 — Cloning my voice for $5 a month. ElevenLabs Starter, a ten-second voice sample, and the feature I came for: per-word timestamps. The thing nobody else gives you on the $5 tier.
Part 4 — Captions that read along with my voice. ASS karaoke format, libass, and the math of pinning subtitles to a fixed mobile anchor so they never jump between scenes. Word-by-word red fill as I read.
Part 5 — Ken Burns is the cheapest motion you can ship. Three SVG layers per scene rendered through Sharp at 2160×3840, animated with FFmpeg’s zoompan, and fused under the captions. The motion baseline before AI got involved.
Part 6 — $0 AI reels on a 12GB GPU. The finale. I swapped the Ken Burns slide deck for real per-scene i2v animation. Wan 2.2 14B plus a Lightning LoRA, running on a home PC over Tailscale. Zero marginal cost per reel.
If you don't have the GPU, the post also walks the cloud equivalent — Replicate Wan 2.2 i2v at ~$0.30 per scene — so the pipeline still ships.
The end state
Take any article you’ve already written and click one button to get a publish-ready vertical video — narrated by you, captioned, branded, and animated. Not a generic stock-footage slideshow. The actual thing.
Total cost at my volume (10–20 reels/month): about $5 in ElevenLabs, $0 in GPU rental, and the electricity my desktop already uses sitting idle. Another $0 build from the workshop.
What I’ll share vs. ship as a kit
I’m not gating the series. The architecture, the decisions, the bug arcs, the cost math, and enough code to understand every stage — all free. By the end of part 6, you’ll be able to replicate the pipeline yourself. It’ll probably take you about twenty hours.
If twenty hours of debugging is not what you want, the kit collapses that to two:
Runnable starter repo. Clone, copy
.env.example,docker compose up. Working pipeline locally in under ten minutes.Three ComfyUI workflow JSONs. My exact Wan 2.2 + Lightning setup, an 8GB-VRAM fallback for older cards, and an image-only Ken Burns workflow for people without a GPU.
Notion playbook. Decision matrices for Wan vs. Ken Burns, voice settings, and caption styles. Tested prompts for the scenario stage. The top ten errors I hit, with fixes.
One Loom walkthrough. Five minutes of me running an article through the pipeline end-to-end, narrating where each piece lives.
Version-pinned README. Tested with: Wan 2.2 14B fp8, ComfyUI v0.X, ElevenLabs v1, FFmpeg 6.1. 12 months of version-pinned updates; after that, you keep the kit as-is.
The kit drops the week after part 6, $39, with a 48-hour early-bird discount for subscribers ($29). Subscribe below, and you’ll get the discount email automatically.
The honest deal
If you read all six posts and decide you’d rather build it yourself, that’s the whole point. The posts are written so that it’s actually possible. The kit exists for people whose Tuesday evenings are worth more than the time it takes to debug FFmpeg pixel formats.
Both are fine outcomes. I’d rather you ship something than buy something.
Part 1 lands next Wednesday. Bookmark this page — I’ll update it with links as each post goes live.
If “one button, one MP4” is the kind of pipeline you’d build yourself, restack with the one tool you keep building because no SaaS does it your way.




Ok, this was May 15th... I assume all 6 parts are done? I want the kit, lol.
I'll head over to the publication to see if it's available. I love your branding - and love that the rebrand tells such a great story of who you are and what you do!