Reelwright

From a line to a short.

Reelwright is a browser studio that turns a topic into a finished vertical video: a scene-bound script, a matching still for every spoken beat, timed voice, karaoke captions, a music bed, and a downloadable file for TikTok, Reels, or Shorts.

It is a response to MoneyPrinterTurbo. That project is a brilliant assembly line. Its ceiling is generic stock. Footage is searched for the topic, not the line, so the picture often lies about the sentence. Reelwright is the director’s cut of the same pipeline.

What you get

1. Type a topic — or paste your own script, one scene per line.
2. Draft locks a hook / point / proof / close before any shots are spent.
3. Produce casts a voice, shoots a still for each line, and cuts the reel.
4. Play it in a phone-shaped stage. Reshoot one scene. Revoice after an edit without regenerating images.
5. Download a WebM/MP4 via the browser recorder.

A sample cut (octopus: three hearts, blue blood) plays on first load so the product is visible before you spend any API calls.

Why this instead of stock-footage printers

MoneyPrinterTurbo	Reelwright
Pexels / Pixabay / Coverr for the whole topic	One still generated for that spoken line
Python, ffmpeg, Whisper, Docker	Browser. No local encoder.
All-or-nothing generate	Draft → edit → produce; revoice; reshoot one shot
Streamlit form	Phone-first cutting room
Random BGM folder	Quiet procedural beds (Pulse / Air / Silent)

The honest limit: stills plus Ken Burns are not live-action clips. They match. Stock often doesn’t.

Stack

• React 19, TanStack Start, Zustand, Tailwind v4
• xAI for the writer (grok-4.5), Imagine stills (grok-imagine-image-2.0), and TTS
• Canvas compositor: cover crop, camera moves, grain, karaoke, progress bar
• Web Audio music bed mixed into MediaRecorder on export
• Brief and recent topics in localStorage — no accounts, no database

Quick start

git clone https://github.com/anwhelan01/reelwright.git
cd reelwright
cp .env.example .env
# put your xAI key in .env as XAI_API_KEY
npm install
npm run dev

Open the URL Vite prints. Without a key the sample cut still plays; Draft and Produce degrade with a clear “AI unavailable” state.

Scripts

Command	What
npm run dev	Dev server
npm run build	Production build
npm run preview	Serve the build
npm run typecheck	tsc --noEmit
npm test	Script unit tests

Environment

Variable	Where	Notes
XAI_API_KEY	server only	Writer, TTS, stills. Never VITE_-prefix it.

Every Produce spends the key owner’s quota: 1 chat + 1 TTS + 4–6 images. Calls are user-initiated. Changing topic/style/duration redrafts; changing format reshoots; editing narration and hitting Revoice skips images.

Using the studio

• 15s / 30s — 4 or 6 scenes. Word budget is written for ~2.6 words/sec.
• 9:16 / 1:1 / 16:9 — TikTok·Reels·Shorts, feed square, landscape.
• Styles — Explainer, List, Story, Fact drop, How-to, Hot take.
• Voices — Orion, Eve, Helix, Luna, Perseus, Sirius, Altair, Rex.
• Captions — Karaoke (word window), Punch (hit words), Lower third.
• Music — Silent, Pulse, Air. Mixed under the voice on play and on export.
• Write the script — one scene per line; needs at least a hook, a point, and a close.
• Space plays/pauses when focus is not in an input.

Pipeline

topic or pasted lines
        │
        ▼
   draftScript (Grok JSON)
        │
        ├─ edit narration / caption in the cut list
        │
        ▼
   produceVoice (TTS + timestamps)
        │
        ▼
   produceSceneImage × N  (parallel, 2 at a time)
        │
        ▼
   canvas cut  →  play  →  MediaRecorder download

Details: docs/ARCHITECTURE.md, docs/PIPELINE.md.

Project layout

src/
  components/studio/   phone stage, cut list, composer
  lib/generate.ts      xAI chat / TTS / images (server functions)
  lib/pipeline.ts      draft → voice → shoot orchestration
  lib/player.ts        canvas compositor
  lib/captions.ts      timestamp alignment
  lib/music.ts         Web Audio beds
  lib/store.ts         Zustand + localStorage
  lib/demo.ts          sample cut
public/demo/           sample stills + voice

Credits

Inspired by harry0703/MoneyPrinterTurbo (MIT) and the original MoneyPrinter. Built with Grok.

License

MIT. See LICENSE.

---

Hermes Bot Mode

This desk is a named Hermes Bot — own model slot, memory, skills, routines, and @mentions.

# once
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash

# this repo
./scripts/install-hermes-bot.sh
hermes -p reelwright chat

In Hermes Desktop the Bot lands under Bots. Type @reelwright from any chat; group it with the rest of the k3ss roster (studio).

File	Role
AGENTS.md	Project harness Hermes loads at session start
.hermes/SOUL.md	Bot personality
.hermes/bot.yaml	Roster, skills, groups
.hermes/skills/reelwright/SKILL.md	Portable skill

Docs: Bot Mode · Context files · Skills