lacy 1.8.11 → 1.8.13

package/docs/UGC_VIDEO_SCRIPT.md

@@ -0,0 +1,114 @@
+# UGC Video Script — Lacy Shell
+
+**Format:** Vertical (9:16), 45-60 seconds
+**Platform:** TikTok, Instagram Reels, YouTube Shorts
+**Style:** Developer at desk, talking to camera + screen recording overlay
+**Tone:** Genuine discovery, casual excitement — not polished ad
+
+---
+
+## Script A: "The Hook" (45 seconds)
+
+### Shot List
+
+| Time | Visual | Audio / Voiceover |
+|------|--------|-------------------|
+| 0-3s | Face to camera, mid-sentence energy | "Okay so I just found this shell plugin and it's kind of insane." |
+| 3-8s | Screen: terminal open, type `git status` — green indicator appears | "You type a normal command — green bar means it runs in your shell like normal." |
+| 8-13s | Screen: clear line, type `why is the build failing` — magenta indicator | "But then you type a question — it turns purple and goes straight to your AI agent." |
+| 13-17s | Screen: hit enter, AI response starts streaming | "No slash commands. No switching apps. It just... figures it out." |
+| 17-22s | Face to camera | "It works with Claude Code, Codex, Gemini — whatever you already use. It's not replacing anything." |
+| 22-28s | Screen: type `npm install`, green indicator. Backspace. Type `install the auth package`, magenta | "Watch — 'npm install' is green, shell command. But 'install the auth package' is purple — same word, different intent." |
+| 28-33s | Screen: type `make sure the tests pass`, shell tries `make`, fails, silently reroutes to AI | "And this is wild — if you accidentally run something that looks like a command but isn't, it catches the error and sends it to AI automatically." |
+| 33-38s | Face to camera, genuine | "The install is literally one line. Curl, npx, or brew. Thirty seconds." |
+| 38-42s | Screen: `curl -fsSL https://lacy.sh/install | bash` | (screen speaks for itself, maybe light background music) |
+| 42-45s | Face to camera | "lacy dot sh. Link in bio." |
+
+### Key Beats
+- **Hook (0-3s):** Mid-sentence, high energy, "I just found this"
+- **Demo (3-33s):** Three distinct wow moments (indicator, same-word routing, auto-reroute)
+- **CTA (33-45s):** Install simplicity + URL
+
+---
+
+## Script B: "The Problem" (50 seconds)
+
+### Shot List
+
+| Time | Visual | Audio / Voiceover |
+|------|--------|-------------------|
+| 0-4s | Face to camera, frustrated energy | "Why do I have to leave my terminal to talk to AI? Every. Single. Time." |
+| 4-8s | Screen: show typical flow — run command, copy output, switch to AI chat, paste, wait | (sped up) "Copy output, switch to Claude, paste, wait, copy the answer, switch back..." |
+| 8-11s | Face to camera | "I found something that fixes this. It's called Lacy." |
+| 11-18s | Screen: terminal with Lacy. Type `ls -la` (green), then type `what does the src folder contain` (magenta) | "You just type. Commands go to your shell. Questions go to your AI. No prefix, no mode switching." |
+| 18-24s | Screen: hit enter on the question, AI responds in-terminal | "The AI responds right there. Same terminal. Same session." |
+| 24-30s | Screen: Ctrl+Space to toggle mode, badge changes | "Ctrl+Space toggles between shell mode, agent mode, and auto mode. Auto figures it out for you." |
+| 30-37s | Screen: type `fix the failing test in auth.test.ts` — magenta indicator, enter, AI starts working | "Just tell it what to do. It routes to whatever AI tool you have — Claude, Codex, Gemini, whatever." |
+| 37-42s | Face to camera | "Open source. Free. Works with ZSH and Bash." |
+| 42-46s | Screen: install command running | "One line to install." |
+| 46-50s | Face to camera, smile | "lacy dot sh." |
+
+### Key Beats
+- **Hook (0-4s):** Relatable frustration — the copy-paste loop
+- **Solution (8-37s):** Show don't tell — actual terminal usage
+- **CTA (37-50s):** Open source credibility + URL
+
+---
+
+## Script C: "The Speed Run" (30 seconds)
+
+For maximum shareability — shortest format.
+
+| Time | Visual | Audio / Voiceover |
+|------|--------|-------------------|
+| 0-3s | Face to camera, fast | "One plugin. Your terminal now speaks English." |
+| 3-7s | Screen: type `git diff` — green indicator | "Commands stay commands." |
+| 7-12s | Screen: type `explain what changed in the last commit` — magenta | "English goes to AI." |
+| 12-16s | Screen: enter, AI responds | "No setup. No prefixes. It just knows." |
+| 16-22s | Screen: rapid montage — `npm test` (green), `why did the test fail` (magenta), `docker ps` (green), `find the memory leak` (magenta) | (fast cuts, maybe trending audio) |
+| 22-26s | Screen: install command | "One line install." |
+| 26-30s | Face to camera | "lacy dot sh. Free. Open source." |
+
+---
+
+## Recording Guidelines
+
+### Setup
+- Clean desk, natural lighting, developer environment visible (second monitor, mechanical keyboard — authentic)
+- Terminal should use a dark theme where green/magenta indicators pop
+- Font size 16-18pt minimum so text is readable on mobile
+- Real project directory (not empty folder) — viewers notice
+
+### Performance Tips
+- Type at 60% your normal speed — viewers need to read the commands
+- Pause 0.5s after each indicator color change so it registers visually
+- Don't rehearse too much — slight stumbles feel authentic for UGC
+- Look directly at camera for face shots
+- Use natural hand gestures when excited about features
+
+### Technical Setup
+- Record terminal separately at 1080p or higher, composite in edit
+- Or use Screen Studio / OBS for picture-in-picture
+- Trending audio optional — works without it for dev audience
+- Add subtle captions (auto-caption tools work fine)
+
+### Variants to A/B Test
+1. **With face** vs **screen-only** (talking over screen recording)
+2. **Problem-first hook** (Script B) vs **discovery hook** (Script A)
+3. **With trending audio** vs **voice only**
+4. **30s version** (Script C) vs **45s version** (Script A)
+
+---
+
+## Hashtags
+
+```
+#developer #coding #terminal #ai #devtools #programming #tech #opensource #productivity #webdev #softwareengineer #claudeai #shell #zsh
+```
+
+## Posting Strategy
+
+- Post 2-3 variants across platforms over 1 week
+- Best times: Tue-Thu, 9-11am or 7-9pm (dev audience)
+- Cross-post to Twitter/X with the screen recording portion as a native video
+- Pin the best-performing version

package/docs/articles/devto-how-i-made-my-terminal-understand-english.md

@@ -0,0 +1,117 @@
+---
+title: How I Made My Terminal Understand English
+published: false
+tags: terminal, ai, devtools, opensource
+canonical_url: https://lacy.sh
+cover_image: https://raw.githubusercontent.com/lacymorrow/lacy/main/docs/demo-full.gif
+---
+
+Every time I need AI help while coding, I do the same thing:
+
+1. Copy terminal output
+2. Switch to Claude or ChatGPT
+3. Paste and ask my question
+4. Wait for the response
+5. Copy the answer
+6. Switch back to terminal
+7. Paste and run
+
+That loop happens 20+ times a day.
+
+AI coding tools like Claude Code and Gemini CLI already live in the terminal. But you still switch into them, type `claude`, ask your thing, then switch back to your regular shell. Two workflows, one terminal.
+
+I built [Lacy Shell](https://lacy.sh) to fix that. It's a ZSH/Bash plugin that figures out whether you're typing a command or a question and routes it to the right place. Commands run in your shell. Questions go to your AI agent. You just type.
+
+![Demo of Lacy Shell showing real-time color indicator](https://raw.githubusercontent.com/lacymorrow/lacy/main/docs/demo-full.gif)
+
+## The color indicator
+
+The thing that made it click for me was the color indicator. As you type, a dot next to your prompt changes color:
+
+- Green = shell command
+- Magenta = AI agent
+
+The first word gets syntax-highlighted too. Both update on every keystroke, so you know what's going to happen before you press Enter.
+
+ZSH also gets a mode badge in the right prompt: `SHELL`, `AGENT`, or `AUTO`. `Ctrl+Space` toggles between them.
+
+## How detection works
+
+The part that surprised me: you don't need AI to classify input. Lacy uses lexical analysis — no network call, no API key. Under a millisecond.
+
+The rules, in priority order:
+
+First, about 150 conversational words ("explain", "why", "thanks", "perfect") always route to AI. `explain this error` has no ambiguity. It's a lookup table, not a model.
+
+Second, shell reserved words. This one tripped me up for a while. Words like `do`, `then`, `in`, and `select` pass `command -v` (they're valid shell syntax), but nobody types `do` and means a `do` loop. "Do we have a way to deploy?" is a question. "In the codebase, where is auth?" is a question. I maintain a list of these and route them straight to AI.
+
+Third, if the first word is a valid command, shell gets it. `git status`, `ls -la`, `docker ps` — straightforward.
+
+Fourth, single non-command words go to shell too. If you type `gti` (a typo for `git`), you probably want to see the shell's error, not have AI interpret it.
+
+Fifth, multiple words where the first isn't a command go to AI. "Fix the bug in auth" starts with "fix" — not a command on most systems. Multiple words, first word isn't a command. That's natural language.
+
+The last rule is my favorite. Sometimes a valid command gets natural language arguments: `kill the process on localhost:3000`. `kill` is a real command, so Lacy sends it to the shell. It fails. Lacy then checks: did the error match a known pattern ("No such process")? Did the input have natural language markers (articles, pronouns, 3+ bare words)? If both, it silently reroutes to the AI agent. You never see the failed attempt.
+
+## Routing table
+
+In practice:
+
+| You type | Routes to | Why |
+|----------|-----------|-----|
+| `ls -la` | Shell | Valid command |
+| `what files are here` | AI | Agent word "what" |
+| `git status` | Shell | Valid command |
+| `do we have auth?` | AI | Reserved word "do" |
+| `cd..` | Shell | Single word (typo) |
+| `fix the bug` | AI | Multi-word, not a command |
+| `kill the process on 3000` | Shell, then AI | Valid command fails with NL patterns |
+| `make sure the tests pass` | Shell, then AI | "sure" is an NL marker, make fails |
+
+## It's a routing layer, not another AI tool
+
+Lacy calls whatever CLI you already have installed:
+
+| Tool | How Lacy calls it |
+|------|------------------|
+| Claude Code | `claude -p "query"` |
+| Gemini CLI | `gemini --resume -p "query"` |
+| OpenCode | `opencode run -c "query"` |
+| Codex | `codex exec resume --last "query"` |
+| Lash | `lash run -c "query"` |
+
+It auto-detects what's on your system, or you can point it at a custom command.
+
+## What I didn't expect
+
+I built this to stop context-switching. The thing that actually changed my workflow was different — I started asking my terminal stuff I'd never have bothered looking up.
+
+What's the flag for recursive grep again? How do I find processes on port 8080? What's the git command to undo the last commit without losing changes?
+
+Before, I'd Google these or fumble through `man` pages. Now I just type the question where I'm already working. The friction was low enough that it changed the habit.
+
+## Install
+
+One line:
+
+```bash
+curl -fsSL https://lacy.sh/install | bash
+```
+
+Or Homebrew:
+
+```bash
+brew install lacymorrow/tap/lacy
+```
+
+Or npx:
+
+```bash
+npx lacy
+```
+
+macOS, Linux, WSL. ZSH and Bash 4+. MIT licensed.
+
+[GitHub](https://github.com/lacymorrow/lacy) / [lacy.sh](https://lacy.sh)
+
+If you try it, I want to hear about the edge cases. The boundary between "command" and "question" is fuzzy by nature, and real usage is how I tune it. File an issue or just yell at me on Twitter.

package/docs/launch-thread-may6.sh

@@ -0,0 +1,158 @@
+#!/usr/bin/env bash
+# LAC-423: Coordinated Twitter launch thread
+# Schedule: May 6, 2026 at 11am ET (15:00 UTC)
+# Account: @lacybuilds via xurl
+# Prerequisite: Show HN posted at 9am ET (LAC-37)
+
+set -euo pipefail
+
+XURL="xurl --app agents -u lacybuilds --auth oauth2"
+GIF_URL="https://raw.githubusercontent.com/lacymorrow/lacy/main/docs/demo-full.gif"
+GIF_PATH="/tmp/lacy-demo-full.gif"
+
+echo "=== LAC-423: Posting Twitter launch thread ==="
+
+# Pre-flight: verify xurl auth
+echo "Checking @lacybuilds auth..."
+AUTH_CHECK=$($XURL whoami 2>&1)
+if echo "$AUTH_CHECK" | grep -q "Unauthorized"; then
+  echo "FATAL: @lacybuilds OAuth2 token expired."
+  echo "Fix: run 'xurl auth oauth2 --app agents' and complete browser flow for lacybuilds."
+  exit 1
+fi
+echo "Auth OK: $(echo "$AUTH_CHECK" | python3 -c "import sys,json; print(json.load(sys.stdin).get('data',{}).get('username','unknown'))" 2>/dev/null || echo 'verified')"
+
+# Step 1: Find the Show HN link
+echo "Searching for Show HN post..."
+HN_LINK=""
+# Search HN for the post (posted ~2 hours earlier)
+HN_SEARCH=$(curl -s "https://hn.algolia.com/api/v1/search_by_date?query=Lacy+Shell&tags=show_hn&hitsPerPage=5" 2>/dev/null || true)
+if [ -n "$HN_SEARCH" ]; then
+  HN_ID=$(echo "$HN_SEARCH" | python3 -c "import sys,json; hits=json.load(sys.stdin).get('hits',[]); print(hits[0]['objectID'] if hits else '')" 2>/dev/null || true)
+  if [ -n "$HN_ID" ]; then
+    HN_LINK="https://news.ycombinator.com/item?id=${HN_ID}"
+    echo "Found HN link: $HN_LINK"
+  fi
+fi
+
+if [ -z "$HN_LINK" ]; then
+  echo "WARNING: Could not find Show HN link. Posting without it."
+  HN_LINK=""
+fi
+
+# Step 2: Download and upload demo GIF
+echo "Downloading demo GIF..."
+curl -fsSL "$GIF_URL" -o "$GIF_PATH"
+
+echo "Uploading media..."
+MEDIA_RESULT=$($XURL media upload "$GIF_PATH" 2>&1)
+MEDIA_ID=$(echo "$MEDIA_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+
+if [ -z "$MEDIA_ID" ]; then
+  echo "WARNING: Media upload failed. Posting without GIF."
+  echo "Upload output: $MEDIA_RESULT"
+fi
+
+# Step 3: Post Tweet 1 (Hook) with GIF
+echo "Posting Tweet 1 (Hook)..."
+TWEET1_TEXT='I made my terminal understand English.
+
+Type a command → it runs in your shell.
+Type a question → it goes to your AI agent.
+
+No prefix. No hotkey. Just type.
+
+It'"'"'s called Lacy Shell. Free and open source.'
+
+if [ -n "$MEDIA_ID" ]; then
+  TWEET1_RESULT=$($XURL post "$TWEET1_TEXT" --media-id "$MEDIA_ID" 2>&1)
+else
+  TWEET1_RESULT=$($XURL post "$TWEET1_TEXT" 2>&1)
+fi
+TWEET1_ID=$(echo "$TWEET1_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+echo "Tweet 1 posted: $TWEET1_ID"
+sleep 3
+
+# Step 4: Post Tweet 2 (Problem) as reply
+echo "Posting Tweet 2 (Problem)..."
+TWEET2_TEXT='The problem: every time you need AI help, you leave your terminal.
+
+Copy output. Switch to Claude/ChatGPT. Paste. Wait. Copy answer. Switch back. Paste.
+
+I was doing that 20+ times a day. So I fixed it.'
+
+TWEET2_RESULT=$($XURL reply "$TWEET1_ID" "$TWEET2_TEXT" 2>&1)
+TWEET2_ID=$(echo "$TWEET2_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+echo "Tweet 2 posted: $TWEET2_ID"
+sleep 3
+
+# Step 5: Post Tweet 3 (How it works) as reply
+echo "Posting Tweet 3 (How it works)..."
+TWEET3_TEXT='How it works:
+
+A color indicator next to your prompt changes as you type:
+
+🟢 Green = runs in your shell
+🟣 Magenta = goes to AI
+
+No AI call to classify your input. Pure lexical analysis. Sub-millisecond.
+
+Command fails with NL patterns? Silently reroutes to AI.'
+
+TWEET3_RESULT=$($XURL reply "$TWEET2_ID" "$TWEET3_TEXT" 2>&1)
+TWEET3_ID=$(echo "$TWEET3_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+echo "Tweet 3 posted: $TWEET3_ID"
+sleep 3
+
+# Step 6: Post Tweet 4 (Tool agnostic) as reply
+echo "Posting Tweet 4 (Tool agnostic)..."
+TWEET4_TEXT='Lacy works with whatever AI tool you already use.
+
+- Claude Code
+- Gemini CLI
+- OpenCode
+- Codex CLI
+- Lash
+- Any custom command
+
+Auto-detects what'"'"'s installed. Zero config.'
+
+TWEET4_RESULT=$($XURL reply "$TWEET3_ID" "$TWEET4_TEXT" 2>&1)
+TWEET4_ID=$(echo "$TWEET4_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+echo "Tweet 4 posted: $TWEET4_ID"
+sleep 3
+
+# Step 7: Post Tweet 5 (CTA) as reply
+echo "Posting Tweet 5 (CTA)..."
+TWEET5_TEXT="One line to install:
+
+curl -fsSL https://lacy.sh/install | bash
+
+Or: brew install lacymorrow/tap/lacy
+
+ZSH + Bash 4+. macOS, Linux, WSL.
+
+github.com/lacymorrow/lacy"
+
+# Append HN link if found
+if [ -n "$HN_LINK" ]; then
+  TWEET5_TEXT="${TWEET5_TEXT}
+
+Live on HN: ${HN_LINK}"
+fi
+
+TWEET5_TEXT="${TWEET5_TEXT}
+
+cc @AnthropicAI @GoogleDeepMind"
+
+TWEET5_RESULT=$($XURL reply "$TWEET4_ID" "$TWEET5_TEXT" 2>&1)
+TWEET5_ID=$(echo "$TWEET5_RESULT" | grep -oE '[0-9]{15,}' | head -1)
+echo "Tweet 5 posted: $TWEET5_ID"
+
+# Done
+echo ""
+echo "=== Thread posted successfully ==="
+echo "Thread URL: https://x.com/lacybuilds/status/$TWEET1_ID"
+
+# Cleanup
+rm -f "$GIF_PATH"

package/docs/videos/README.md

@@ -0,0 +1,189 @@
+# UGC Video Generation
+
+Programmatically generated UGC (User-Generated Content) demo videos for Lacy Shell marketing.
+
+## Videos
+
+### V2 (current — enhanced production value)
+
+| File | Duration | Format | Description |
+|------|----------|--------|-------------|
+| `lacy-shell-demo-v2.mp4` | 28s | 1080x1920 **60fps** H.264 | Full demo — 6 scenes, gradient BG, glow effects, drop shadows |
+| `lacy-shell-short-v2.mp4` | 15s | 1080x1920 **60fps** H.264 | Short — hook + demo + CTA, same V2 enhancements |
+
+### V1 (archived)
+
+| File | Duration | Format | Description |
+|------|----------|--------|-------------|
+| `lacy-shell-demo.mp4` | 27.5s | 1080x1920 30fps H.264 | Full demo — 6 scenes covering all key features |
+| `lacy-shell-short.mp4` | 15s | 1080x1920 30fps H.264 | Short version — hook + demo + CTA for TikTok/Reels |
+
+Both formats are vertical (9:16), optimized for TikTok, Instagram Reels, and YouTube Shorts.
+
+## How It Works
+
+Videos are rendered frame-by-frame using Python (Pillow for image generation) and assembled into MP4 with ffmpeg. No external services or accounts needed.
+
+**Pipeline:**
+1. Python script generates individual PNG frames with animations (typing, fades, easing, glow compositing)
+2. ffmpeg encodes frames into H.264 MP4
+
+## V2 Visual Enhancements
+
+| Feature | Detail |
+|---------|--------|
+| **60fps** | Doubled frame rate for premium smoothness |
+| **Gradient background** | Deep purple-black gradient instead of flat black |
+| **Glow/bloom on indicators** | GaussianBlur compositing creates soft halo on green/magenta dots |
+| **Drop shadows** | Blurred shadow beneath every terminal window |
+| **CRT scan-lines** | Subtle 3px scan-line overlay for retro terminal feel |
+| **Variable typing speed** | Eased typing with natural acceleration |
+| **Scene fade transitions** | 0.3s fade-in/out at each scene boundary |
+| **Progress bar** | Thin colored bar tracks overall video progress |
+| **Accessibility captions** | Semi-transparent caption pills explain each scene |
+| **Floating particles** | Faint code snippets drift upward in background |
+| **Elastic spring animations** | Key text reveals use overshoot easing for punch |
+| **Thumbnail export** | Best CTA frame auto-saved as PNG for social previews |
+
+## Prerequisites
+
+```bash
+# Python 3.10+ with Pillow
+python3 -m venv .venv
+.venv/bin/pip install Pillow
+
+# ffmpeg
+brew install ffmpeg   # macOS
+```
+
+## Generating Videos
+
+### V2 Full Demo (28s, 60fps)
+
+```bash
+# Generate frames
+.venv/bin/python3 docs/videos/generate_frames_v2.py
+
+# Assemble into MP4
+ffmpeg -framerate 60 -i /tmp/ugc-video/frames-v2/frame_%05d.png \
+  -c:v libx264 -pix_fmt yuv420p -crf 18 \
+  -y docs/videos/lacy-shell-demo-v2.mp4
+```
+
+### V2 Short (15s, 60fps)
+
+```bash
+# Generate frames
+.venv/bin/python3 docs/videos/generate_short_v2.py
+
+# Assemble into MP4
+ffmpeg -framerate 60 -i /tmp/ugc-video/frames-short-v2/frame_%05d.png \
+  -c:v libx264 -pix_fmt yuv420p -crf 18 \
+  -y docs/videos/lacy-shell-short-v2.mp4
+```
+
+### Platform-specific exports
+
+```bash
+# TikTok/Reels/Shorts — 9:16 native, no extra crop needed
+# Use lacy-shell-short-v2.mp4 directly
+
+# Twitter/X — square crop
+ffmpeg -i docs/videos/lacy-shell-demo-v2.mp4 \
+  -vf 'crop=1080:1080:0:420' \
+  -c:v libx264 -pix_fmt yuv420p -crf 20 \
+  -y docs/videos/lacy-shell-demo-v2-square.mp4
+
+# Instagram square (short)
+ffmpeg -i docs/videos/lacy-shell-short-v2.mp4 \
+  -vf 'crop=1080:1080:0:420' \
+  -c:v libx264 -pix_fmt yuv420p -crf 20 \
+  -y docs/videos/lacy-shell-short-v2-square.mp4
+```
+
+### V1 (legacy — 30fps, flat black BG)
+
+```bash
+# Full demo (v1)
+.venv/bin/python3 docs/videos/generate_frames.py
+ffmpeg -framerate 30 -i /tmp/ugc-video/frames/frame_%05d.png \
+  -c:v libx264 -pix_fmt yuv420p -crf 18 \
+  -y docs/videos/lacy-shell-demo.mp4
+
+# Short (v1)
+.venv/bin/python3 docs/videos/generate_short.py
+ffmpeg -framerate 30 -i /tmp/ugc-video/frames-short/frame_%05d.png \
+  -c:v libx264 -pix_fmt yuv420p -crf 18 \
+  -y docs/videos/lacy-shell-short.mp4
+```
+
+## Scene Breakdown
+
+### Full Demo (`generate_frames.py`)
+
+| Scene | Duration | Description |
+|-------|----------|-------------|
+| Hook | 3.5s | "What if your shell understood you?" — word-by-word fade-in |
+| Shell Command | 4.5s | `ls -la` with green indicator, file listing output |
+| NL Query | 5.5s | "what files are in this project" with magenta indicator, AI response |
+| Split Comparison | 5s | "Same word. Different intent." — `install nodejs` (shell) vs `install a way to monitor logs` (agent) |
+| Auto-Reroute | 5.5s | "fix the failing tests" — fails in shell, auto-routes to AI |
+| CTA | 3.5s | lacy.sh branding, install command, platform features |
+
+### Short Version (`generate_short.py`)
+
+| Scene | Duration | Description |
+|-------|----------|-------------|
+| Hook | 2s | "Stop copy-pasting into ChatGPT." |
+| Demo | 8s | `git status` (green/shell) then "explain what changed and why" (magenta/agent) |
+| CTA | 5s | lacy.sh + install command + features |
+
+## Customizing
+
+### Changing text/commands
+
+Edit the scene definitions in the V2 generator scripts. Key areas:
+
+- **`generate_frames_v2.py`**: Each `TerminalTypingScene()` takes `command`, `response_lines`, `label`, `indicator_color`, and `caption`
+- **`generate_short_v2.py`**: Scene functions (`scene_hook`, `scene_demo`, `scene_cta`) contain the text inline
+
+### Changing colors
+
+Color constants are at the top of each script:
+
+```python
+GREEN = (52, 211, 153)      # Shell indicator
+MAGENTA = (216, 100, 240)   # Agent indicator
+BLUE = (96, 165, 250)       # Auto mode badge
+```
+
+### Changing duration
+
+Adjust `duration_sec` on each scene (full demo) or the duration tuple (short version). Frame count = duration * FPS.
+
+### Changing resolution
+
+Modify `W, H` at the top. Current: `1080, 1920` (9:16 vertical). For landscape: `1920, 1080`.
+
+### Font
+
+Scripts use Menlo (macOS default monospace). Falls back to SF NS Mono or system default. To use a different font, update the `load_font()` function paths.
+
+## Output
+
+| Path | Contents |
+|------|----------|
+| `/tmp/ugc-video/frames-v2/` | V2 full demo PNG frames |
+| `/tmp/ugc-video/frames-short-v2/` | V2 short PNG frames |
+| `/tmp/ugc-video/lacy-shell-thumbnail.png` | Full demo CTA thumbnail |
+| `/tmp/ugc-video/lacy-shell-short-thumbnail.png` | Short CTA thumbnail |
+| `/tmp/ugc-video/frames/` | V1 full demo frames (legacy) |
+| `/tmp/ugc-video/frames-short/` | V1 short frames (legacy) |
+
+Final MP4s go to `docs/videos/`.
+
+## Quality Settings
+
+- **CRF 18**: High quality. Lower = better quality, bigger file. Range: 0-51.
+- **60fps (V2)**: Premium smoothness for easing animations and glow effects. Use 30fps to halve file size.
+- **H.264 + yuv420p**: Maximum compatibility across all platforms and devices.