Shorts

{
  "name": "shorts",
  "version": "1.1.0",
  "description": "Interactive longform-to-shortform video creator. Extracts viral-ready short clips from long videos with Remotion-rendered animated captions, AI segment scoring, and platform-optimized exports for YouTube Shorts, TikTok, and Instagram Reels.",
  "author": "AgriciDaniel",
  "license": "MIT",
  "homepage": "https://github.com/AgriciDaniel/claude-shorts"
}

version: 2
updates:
  - package-ecosystem: "pip"
    directory: "/"
    schedule:
      interval: "weekly"
    labels:
      - "dependencies"
    open-pull-requests-limit: 5
  - package-ecosystem: "github-actions"
    directory: "/"
    schedule:
      interval: "weekly"
    labels:
      - "dependencies"

name: Bug Report
description: Report a bug or unexpected behavior
labels: ["bug"]
body:
  - type: markdown
    attributes:
      value: Thanks for reporting a bug! Please fill out the details below.
  - type: textarea
    id: description
    attributes:
      label: What happened?
      description: Describe the bug clearly
    validations:
      required: true
  - type: textarea
    id: expected
    attributes:
      label: Expected behavior
      description: What did you expect to happen?
    validations:
      required: true
  - type: textarea
    id: reproduce
    attributes:
      label: Steps to reproduce
      description: How can we reproduce this?
      placeholder: |
        1. ...
        2. ...
        3. See error
    validations:
      required: true
  - type: input
    id: environment
    attributes:
      label: Environment
      description: OS, runtime version, or any relevant environment details
      placeholder: "e.g., macOS 14, Python 3.12, Node 20"
  - type: textarea
    id: logs
    attributes:
      label: Error output
      description: Paste any error messages or logs
      render: shell

blank_issues_enabled: false
contact_links:
  - name: Discussions
    url: https://github.com/AgriciDaniel/claude-shorts/issues
    about: Ask questions and share ideas

name: Feature Request
description: Suggest a new feature or improvement
labels: ["enhancement"]
body:
  - type: markdown
    attributes:
      value: We'd love to hear your ideas!
  - type: textarea
    id: problem
    attributes:
      label: Problem or use case
      description: What problem does this solve or what would it enable?
    validations:
      required: true
  - type: textarea
    id: solution
    attributes:
      label: Proposed solution
      description: How would you like this to work?
    validations:
      required: true
  - type: dropdown
    id: area
    attributes:
      label: Area
      options:
        - Bug fix
        - New feature
        - Documentation
        - Performance
        - Other

changelog:
  categories:
    - title: New Features
      labels:
        - enhancement
    - title: Bug Fixes
      labels:
        - bug
    - title: Documentation
      labels:
        - documentation
    - title: Dependencies
      labels:
        - dependencies
    - title: Other Changes
      labels:
        - "*"

# Generated outputs
shorts/
/tmp/

# Remotion cache
.remotion/

# Dependencies
node_modules/

# Python
__pycache__/
*.pyc
*.pyo
*.egg-info/
dist/
build/

# Virtual environments
.venv/
venv/

# Environment variables
.env
.env.local
.env.*.local

# OS files
.DS_Store
Thumbs.db
*.swp
*.swo
*~

# IDE
.vscode/
.idea/
*.code-workspace

# Temporary / intermediate files
*.wav
*.tmp

#!/usr/bin/env bash
# Install claude-shorts skill to ~/.claude/skills/
set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
SKILL_DIR="$HOME/.claude/skills/shorts"

echo "=== Installing shorts skill ==="

# Create skill directory
mkdir -p "$SKILL_DIR"
mkdir -p "$SKILL_DIR/scripts"
mkdir -p "$SKILL_DIR/references"
mkdir -p "$SKILL_DIR/remotion"

# Copy SKILL.md (the skill definition)
cp "$SCRIPT_DIR/SKILL.md" "$SKILL_DIR/SKILL.md"

# Copy scripts
for f in "$SCRIPT_DIR"/scripts/*.sh "$SCRIPT_DIR"/scripts/*.py; do
    [ -f "$f" ] && cp "$f" "$SKILL_DIR/scripts/"
done

# Make scripts executable
chmod +x "$SKILL_DIR"/scripts/*.sh 2>/dev/null || true
chmod +x "$SKILL_DIR"/scripts/*.py 2>/dev/null || true

# Copy references
for f in "$SCRIPT_DIR"/references/*.md; do
    [ -f "$f" ] && cp "$f" "$SKILL_DIR/references/"
done

# Copy Remotion project (excluding node_modules)
rsync -a --exclude='node_modules' --exclude='.remotion' \
    "$SCRIPT_DIR/remotion/" "$SKILL_DIR/remotion/"

echo ""
echo "Installed to: $SKILL_DIR"
echo ""
echo "Files copied:"
find "$SKILL_DIR" -type f | sort | while read -r f; do
    echo "  ${f#$SKILL_DIR/}"
done
echo ""
echo "Next steps:"
echo "  1. Run 'bash setup.sh' to install dependencies (if not done)"
echo "  2. Use /shorts in Claude Code to start the interactive pipeline"

MIT License

Copyright (c) 2026 Daniel Agrici

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
SOFTWARE.

{
  "name": "claude-shorts-remotion",
  "version": "1.0.0",
  "private": true,
  "scripts": {
    "preview": "remotion preview src/index.ts",
    "render": "node render.mjs"
  },
  "dependencies": {
    "@remotion/bundler": "^4.0.422",
    "@remotion/captions": "^4.0.422",
    "@remotion/cli": "^4.0.422",
    "@remotion/renderer": "^4.0.422",
    "react": "^19.0.0",
    "react-dom": "^19.0.0",
    "remotion": "^4.0.422",
    "zod": "^3.23.0"
  },
  "devDependencies": {
    "@types/react": "^19.0.0",
    "typescript": "^5.5.0"
  }
}

Copyright 2011 The Montserrat Project Authors (https://github.com/JulietaUla/Montserrat)
Copyright 2010 The Bangers Project Authors (https://github.com/googlefonts/bangers)
Copyright 2016 The Inter Project Authors (https://github.com/rsms/inter)

This Font Software is licensed under the SIL Open Font License, Version 1.1.
This license is copied below, and is also available with a FAQ at:
https://openfontlicense.org


-----------------------------------------------------------
SIL OPEN FONT LICENSE Version 1.1 - 26 February 2007
-----------------------------------------------------------

PREAMBLE
The goals of the Open Font License (OFL) are to stimulate worldwide
development of collaborative font projects, to support the font creation
efforts of academic and linguistic communities, and to provide a free and
open framework in which fonts may be shared and improved in partnership
with others.

The OFL allows the licensed fonts to be used, studied, modified and
redistributed freely as long as they are not sold by themselves. The
fonts, including any derivative works, can be bundled, embedded,
redistributed and/or sold with any software provided that any reserved
names are not used by derivative works. The fonts and derivatives,
however, cannot be released under any other type of license. The
requirement for fonts to remain under this license does not apply
to any document created using the fonts or their derivatives.

DEFINITIONS
"Font Software" refers to the set of files released by the Copyright
Holder(s) under this license and clearly marked as such. This may
include source files, build scripts and documentation.

"Reserved Font Name" refers to any names specified as such after the
copyright statement(s).

"Original Version" refers to the collection of Font Software components as
distributed by the Copyright Holder(s).

"Modified Version" refers to any derivative made by adding to, deleting,
or substituting -- in part or in whole -- any of the components of the
Original Version, by changing formats or by porting the Font Software to a
new environment.

"Author" refers to any designer, engineer, programmer, technical
writer or other person who contributed to the Font Software.

PERMISSION & CONDITIONS
Permission is hereby granted, free of charge, to any person obtaining
a copy of the Font Software, to use, study, copy, merge, embed, modify,
redistribute, and sell modified and unmodified copies of the Font
Software, subject to the following conditions:

1) Neither the Font Software nor any of its individual components,
in Original or Modified Versions, may be sold by itself.

2) Original or Modified Versions of the Font Software may be bundled,
redistributed and/or sold with any software, provided that each copy
contains the above copyright notice and this license. These can be
included either as stand-alone text files, human-readable headers or
in the appropriate machine-readable metadata fields within text or
binary files as long as those fields can be easily viewed by the user.

3) No Modified Version of the Font Software may use the Reserved Font
Name(s) unless explicit written permission is granted by the corresponding
Copyright Holder. This restriction only applies to the primary font name as
presented to the users.

4) The name(s) of the Copyright Holder(s) or the Author(s) of the Font
Software shall not be used to promote, endorse or advertise any
Modified Version, except to acknowledge the contribution(s) of the
Copyright Holder(s) and the Author(s) or with their explicit written
permission.

5) The Font Software, modified or unmodified, in part or in whole,
must be distributed entirely under this license, and must not be
distributed under any other license. The requirement for fonts to
remain under this license does not apply to any document created
using the Font Software.

TERMINATION
This license becomes null and void if any of the above conditions are
not met.

DISCLAIMER
THE FONT SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY WARRANTIES OF
MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT
OF COPYRIGHT, PATENT, TRADEMARK, OR OTHER RIGHT. IN NO EVENT SHALL THE
COPYRIGHT HOLDER BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
INCLUDING ANY GENERAL, SPECIAL, INDIRECT, INCIDENTAL, OR CONSEQUENTIAL
DAMAGES, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
FROM, OUT OF THE USE OR INABILITY TO USE THE FONT SOFTWARE OR FROM
OTHER DEALINGS IN THE FONT SOFTWARE.

import { Config } from "@remotion/cli/config";

Config.setVideoImageFormat("jpeg");
Config.setOverwriteOutput(true);

#!/usr/bin/env node
/**
 * Bundle-once-render-many orchestrator for claude-shorts.
 *
 * Bundles the Remotion project once, opens a shared Chrome instance,
 * then renders each approved segment sequentially.
 *
 * Usage:
 *   node render.mjs \
 *     --segments /tmp/claude-shorts/approved_segments.json \
 *     --reframe /tmp/claude-shorts/reframe.json \
 *     --captions /tmp/claude-shorts/transcript.json \
 *     --style bold \
 *     --clips-dir /tmp/claude-shorts/clips/ \
 *     --output-dir /tmp/claude-shorts/render/
 *
 * IMPORTANT: Uses selectComposition() before renderMedia() to resolve
 * calculateMetadata (dynamic duration per segment).
 */
import { bundle } from "@remotion/bundler";
import { renderMedia, selectComposition, openBrowser } from "@remotion/renderer";
import path from "path";
import fs from "fs";
import http from "http";
import { fileURLToPath } from "url";

const __dirname = path.dirname(fileURLToPath(import.meta.url));

function parseArgs() {
  const args = process.argv.slice(2);
  const parsed = {};
  for (let i = 0; i < args.length; i += 2) {
    const key = args[i].replace(/^--/, "");
    parsed[key] = args[i + 1];
  }
  return parsed;
}

async function main() {
  const args = parseArgs();

  const segmentsPath = args.segments;
  const reframePath = args.reframe;
  const captionsPath = args.captions;
  const style = args.style || "bold";
  const clipsDir = args["clips-dir"];
  const outputDir = args["output-dir"] || "/tmp/claude-shorts/render/";

  if (!segmentsPath || !reframePath || !captionsPath || !clipsDir) {
    console.error(JSON.stringify({
      error: "Missing required arguments",
      usage: "node render.mjs --segments FILE --reframe FILE --captions FILE --clips-dir DIR [--style bold|bounce|clean] [--output-dir DIR]",
    }));
    process.exit(1);
  }

  // Load data files
  const segmentsData = JSON.parse(fs.readFileSync(segmentsPath, "utf-8"));
  const reframeData = JSON.parse(fs.readFileSync(reframePath, "utf-8"));
  const transcriptData = JSON.parse(fs.readFileSync(captionsPath, "utf-8"));

  const segments = segmentsData.segments;
  const allCaptions = transcriptData.captions;

  // Ensure output directory exists
  fs.mkdirSync(outputDir, { recursive: true });

  // Start a local HTTP server to serve clip files.
  // Remotion's renderer proxy only supports http/https URLs, not file://.
  const clipServer = http.createServer((req, res) => {
    const filePath = path.join(clipsDir, decodeURIComponent(req.url.slice(1)));
    if (!fs.existsSync(filePath)) {
      res.writeHead(404);
      res.end("Not found");
      return;
    }
    res.writeHead(200, { "Content-Type": "video/mp4" });
    fs.createReadStream(filePath).pipe(res);
  });

  await new Promise((resolve) => clipServer.listen(0, "127.0.0.1", resolve));
  const clipServerPort = clipServer.address().port;
  const clipBaseUrl = `http://127.0.0.1:${clipServerPort}`;
  console.log(JSON.stringify({ action: "clip_server", port: clipServerPort, url: clipBaseUrl }));

  console.log(JSON.stringify({ action: "bundle", status: "starting" }));
  const startBundle = Date.now();

  // Bundle once
  const serveUrl = await bundle({
    entryPoint: path.resolve(__dirname, "src/index.ts"),
  });

  const bundleTime = ((Date.now() - startBundle) / 1000).toFixed(1);
  console.log(JSON.stringify({ action: "bundle", status: "complete", time_sec: bundleTime }));

  // Open shared browser instance
  const browser = await openBrowser("chrome");

  const results = [];

  try {
    for (let i = 0; i < segments.length; i++) {
      const seg = segments[i];
      const segId = seg.id ?? (i + 1);
      const clipName = `clip_${String(segId).padStart(2, "0")}.mp4`;
      const clipPath = path.resolve(clipsDir, clipName);
      const outputPath = path.resolve(outputDir, `short_${String(segId).padStart(2, "0")}.mp4`);

      if (!fs.existsSync(clipPath)) {
        console.error(JSON.stringify({ error: `Clip not found: ${clipPath}` }));
        continue;
      }

      // Get reframe data for this clip
      const reframe = reframeData.clips?.[clipName] || {};
      if (!reframe.crop) {
        console.error(JSON.stringify({
          warning: `No reframe data for ${clipName}, using default 9:16 center crop for 1920x1080`,
        }));
      }
      const crop = reframe.crop || { x: 0, y: 0, w: 607, h: 1080 };
      const cropKeyframes = reframe.crop_keyframes || [];

      // Get source resolution from reframe data
      const [srcW, srcH] = (reframe.source_resolution || "1920x1080")
        .split("x")
        .map(Number);

      // Extract captions for this segment's time range (in milliseconds)
      const segStartMs = seg.start * 1000;
      const segEndMs = seg.end * 1000;
      const segCaptions = allCaptions
        .filter((c) => c.startMs >= segStartMs && c.endMs <= segEndMs)
        .map((c) => ({
          text: c.text,
          startMs: c.startMs - segStartMs, // Offset to clip-local time
          endMs: c.endMs - segStartMs,
        }));

      const durationInSeconds = seg.end - seg.start;

      const inputProps = {
        clipSrc: `${clipBaseUrl}/${clipName}`,
        sourceWidth: srcW,
        sourceHeight: srcH,
        crop,
        cropKeyframes,
        captions: segCaptions,
        captionStyle: style,
        hookLine1: seg.hook_line1 || "",
        hookLine2: seg.hook_line2 || "",
        showProgressBar: true,
        durationInSeconds,
      };

      console.log(JSON.stringify({
        action: "render",
        segment: i + 1,
        total: segments.length,
        duration: `${durationInSeconds.toFixed(1)}s`,
        status: "starting",
      }));

      // Retry once on failure before skipping this segment
      let rendered = false;
      const maxAttempts = 2;
      for (let attempt = 1; attempt <= maxAttempts; attempt++) {
        try {
          const startRender = Date.now();

          // selectComposition resolves calculateMetadata for dynamic duration
          const composition = await selectComposition({
            serveUrl,
            id: "ShortVideo",
            inputProps,
            puppeteerInstance: browser,
          });

          await renderMedia({
            composition,
            serveUrl,
            codec: "h264",
            crf: 18,
            outputLocation: outputPath,
            inputProps,
            puppeteerInstance: browser,
          });

          const renderTime = ((Date.now() - startRender) / 1000).toFixed(1);
          const fileSizeMB = (fs.statSync(outputPath).size / 1024 / 1024).toFixed(1);

          const result = {
            action: "render",
            segment: i + 1,
            output: outputPath,
            duration: `${durationInSeconds.toFixed(1)}s`,
            render_time_sec: renderTime,
            file_size_mb: fileSizeMB,
            status: "complete",
          };

          results.push(result);
          console.log(JSON.stringify(result));
          rendered = true;
          break;
        } catch (renderErr) {
          if (attempt < maxAttempts) {
            console.log(JSON.stringify({
              action: "render",
              segment: i + 1,
              status: "retrying",
              attempt,
              error: renderErr.message,
            }));
          } else {
            console.error(JSON.stringify({
              action: "render",
              segment: i + 1,
              status: "failed",
              error: renderErr.message,
            }));
          }
        }
      }

      if (!rendered) {
        results.push({
          action: "render",
          segment: i + 1,
          output: outputPath,
          status: "failed",
        });
      }
    }
  } finally {
    await browser.close({ silent: true });
    clipServer.close();
  }

  // Final summary
  console.log(JSON.stringify({
    action: "render_complete",
    segments_rendered: results.length,
    total_segments: segments.length,
    output_dir: outputDir,
    results,
  }));
}

main().catch((err) => {
  console.error(JSON.stringify({ error: err.message, stack: err.stack }));
  process.exit(1);
});

import { useCurrentFrame, useVideoConfig, spring } from "remotion";
import { useCaptionPages } from "../hooks/useCaptionPages";
import { BOLD_THEME } from "../styles/theme";
import type { Caption } from "../types";

interface BoldCaptionsProps {
  captions: Caption[];
}

/**
 * Bold-style captions: ALL CAPS, pop-in spring animation,
 * yellow highlight on the currently active word.
 *
 * Font: Montserrat Bold
 * Words per page: 2-3 (800ms combine window)
 * Animation: Pop-in spring {mass:1, damping:12, stiffness:200}
 */
export const BoldCaptions: React.FC<BoldCaptionsProps> = ({
  captions,
}) => {
  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const currentTimeMs = (frame / fps) * 1000;

  const pages = useCaptionPages(captions, 800);

  // Find the current page
  const currentPage = pages.find(
    (p) =>
      currentTimeMs >= p.startMs &&
      currentTimeMs < p.startMs + p.durationMs
  );

  if (!currentPage) return null;

  const pageStartFrame = Math.floor((currentPage.startMs / 1000) * fps);
  const localFrame = frame - pageStartFrame;

  // Pop-in spring animation
  const scale = spring({
    frame: localFrame,
    fps,
    config: { mass: 1, damping: 12, stiffness: 200 },
  });

  return (
    <div
      style={{
        position: "absolute",
        bottom: 350,
        left: 40,
        right: 40,
        display: "flex",
        justifyContent: "center",
        alignItems: "center",
        flexWrap: "wrap",
        gap: 12,
        transform: `scale(${scale})`,
      }}
    >
      {currentPage.tokens.map((token, i) => {
        const isActive =
          currentTimeMs >= token.fromMs && currentTimeMs < token.toMs;

        return (
          <span
            key={`${currentPage.startMs}-${i}`}
            style={{
              fontFamily: "'Montserrat', sans-serif",
              fontWeight: 800,
              fontSize: 72,
              textTransform: "uppercase",
              color: isActive
                ? BOLD_THEME.activeColor
                : BOLD_THEME.textColor,
              textShadow: `
                3px 3px 0 ${BOLD_THEME.shadowColor},
                -3px -3px 0 ${BOLD_THEME.shadowColor},
                3px -3px 0 ${BOLD_THEME.shadowColor},
                -3px 3px 0 ${BOLD_THEME.shadowColor}
              `,
              lineHeight: 1.1,
              textAlign: "center",
            }}
          >
            {token.text.trim().toUpperCase()}
          </span>
        );
      })}
    </div>
  );
};

import { useCurrentFrame, useVideoConfig, spring } from "remotion";
import { useCaptionPages } from "../hooks/useCaptionPages";
import { BOUNCE_THEME } from "../styles/theme";
import type { Caption } from "../types";

interface BounceCaptionsProps {
  captions: Caption[];
}

/**
 * Bounce-style captions: Bouncy scale spring, rotating bright colors,
 * 1-2 words per page for maximum impact.
 *
 * Font: Bangers (Google Fonts, OFL)
 * Words per page: 1-2 (600ms combine window)
 * Animation: Scale spring 70% → 120% → 100% {mass:1, damping:8}
 */
export const BounceCaptions: React.FC<BounceCaptionsProps> = ({
  captions,
}) => {
  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const currentTimeMs = (frame / fps) * 1000;

  const pages = useCaptionPages(captions, 600);

  const currentPage = pages.find(
    (p) =>
      currentTimeMs >= p.startMs &&
      currentTimeMs < p.startMs + p.durationMs
  );

  if (!currentPage) return null;

  const pageStartFrame = Math.floor((currentPage.startMs / 1000) * fps);
  const localFrame = frame - pageStartFrame;
  const pageIndex = pages.indexOf(currentPage);

  // Bouncy scale spring: starts at 0.7, overshoots to ~1.2, settles at 1.0
  const rawScale = spring({
    frame: localFrame,
    fps,
    config: { mass: 1, damping: 8, stiffness: 180 },
  });
  const scale = 0.7 + rawScale * 0.3;

  // Rotating color per page
  const color = BOUNCE_THEME.rotatingColors[pageIndex % BOUNCE_THEME.rotatingColors.length];

  return (
    <div
      style={{
        position: "absolute",
        bottom: 350,
        left: 40,
        right: 40,
        display: "flex",
        justifyContent: "center",
        alignItems: "center",
        transform: `scale(${scale})`,
      }}
    >
      <span
        style={{
          fontFamily: "'Bangers', cursive",
          fontSize: 84,
          color,
          textShadow: `
            4px 4px 0 ${BOUNCE_THEME.shadowColor},
            -4px -4px 0 ${BOUNCE_THEME.shadowColor},
            4px -4px 0 ${BOUNCE_THEME.shadowColor},
            -4px 4px 0 ${BOUNCE_THEME.shadowColor}
          `,
          textAlign: "center",
          lineHeight: 1.0,
          textTransform: "uppercase",
        }}
      >
        {currentPage.text.trim().toUpperCase()}
      </span>
    </div>
  );
};

import type { Caption, CaptionStyleType } from "../types";
import { BoldCaptions } from "./BoldCaptions";
import { BounceCaptions } from "./BounceCaptions";
import { CleanCaptions } from "./CleanCaptions";

interface CaptionsProps {
  captions: Caption[];
  style: CaptionStyleType;
}

/**
 * Style dispatcher — routes to the correct caption renderer.
 */
export const Captions: React.FC<CaptionsProps> = ({ captions, style }) => {
  if (!captions || captions.length === 0) return null;

  switch (style) {
    case "bold":
      return <BoldCaptions captions={captions} />;
    case "bounce":
      return <BounceCaptions captions={captions} />;
    case "clean":
      return <CleanCaptions captions={captions} />;
    default:
      return <BoldCaptions captions={captions} />;
  }
};

import { useCurrentFrame, useVideoConfig, interpolate } from "remotion";
import { useCaptionPages } from "../hooks/useCaptionPages";
import { CLEAN_THEME } from "../styles/theme";
import type { Caption } from "../types";

interface CleanCaptionsProps {
  captions: Caption[];
}

/**
 * Clean-style captions: Minimal fade-in, white text with subtle shadow,
 * 3-5 words per page for readability.
 *
 * Font: Inter Bold
 * Words per page: 3-5 (1500ms combine window)
 * Animation: Fade-in opacity 0 → 1 over 6 frames
 */
export const CleanCaptions: React.FC<CleanCaptionsProps> = ({ captions }) => {
  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const currentTimeMs = (frame / fps) * 1000;

  const pages = useCaptionPages(captions, 1500);

  const currentPage = pages.find(
    (p) =>
      currentTimeMs >= p.startMs &&
      currentTimeMs < p.startMs + p.durationMs
  );

  if (!currentPage) return null;

  const pageStartFrame = Math.floor((currentPage.startMs / 1000) * fps);
  const localFrame = frame - pageStartFrame;

  // Fade-in over 6 frames
  const opacity = interpolate(localFrame, [0, 6], [0, 1], {
    extrapolateRight: "clamp",
  });

  return (
    <div
      style={{
        position: "absolute",
        bottom: 350,
        left: 60,
        right: 60,
        display: "flex",
        justifyContent: "center",
        alignItems: "center",
        opacity,
      }}
    >
      {currentPage.tokens.map((token, i) => {
        const isActive =
          currentTimeMs >= token.fromMs && currentTimeMs < token.toMs;

        return (
          <span
            key={`${currentPage.startMs}-${i}`}
            style={{
              fontFamily: "'Inter', sans-serif",
              fontWeight: 700,
              fontSize: 56,
              color: isActive
                ? CLEAN_THEME.activeColor
                : CLEAN_THEME.textColor,
              textShadow: `2px 2px 8px ${CLEAN_THEME.shadowColor}`,
              lineHeight: 1.3,
              textAlign: "center",
              marginRight: 8,
            }}
          >
            {token.text.trim()}
          </span>
        );
      })}
    </div>
  );
};

import { useCurrentFrame, useVideoConfig, spring, interpolate } from "remotion";

interface HookOverlayProps {
  line1: string;
  line2: string;
}

/**
 * Hook text overlay shown in the first 3.5 seconds.
 * Line 1: Large white text (main hook)
 * Line 2: Smaller cyan text (subtitle/context)
 *
 * Animation: Spring pop-in at 0.3s, fade-out at 3.0s
 */
export const HookOverlay: React.FC<HookOverlayProps> = ({ line1, line2 }) => {
  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const currentTimeSec = frame / fps;

  // Only show between 0.3s and 3.5s
  if (currentTimeSec < 0.3 || currentTimeSec > 3.5) return null;

  const enterFrame = Math.max(0, frame - Math.floor(0.3 * fps));

  // Pop-in spring
  const scale = spring({
    frame: enterFrame,
    fps,
    config: { mass: 1, damping: 14, stiffness: 200 },
  });

  // Fade-out starting at 3.0s
  const opacity = interpolate(
    currentTimeSec,
    [3.0, 3.5],
    [1, 0],
    { extrapolateLeft: "clamp", extrapolateRight: "clamp" }
  );

  return (
    <div
      style={{
        position: "absolute",
        top: 40,
        left: 40,
        right: 40,
        display: "flex",
        flexDirection: "column",
        alignItems: "center",
        gap: 8,
        opacity,
        transform: `scale(${scale})`,
      }}
    >
      {line1 && (
        <div
          style={{
            fontFamily: "'Montserrat', sans-serif",
            fontWeight: 800,
            fontSize: 48,
            color: "white",
            textShadow: "3px 3px 0 black, -3px -3px 0 black, 3px -3px 0 black, -3px 3px 0 black",
            textAlign: "center",
            lineHeight: 1.2,
          }}
        >
          {line1}
        </div>
      )}
      {line2 && (
        <div
          style={{
            fontFamily: "'Inter', sans-serif",
            fontWeight: 600,
            fontSize: 28,
            color: "#00BFFF",
            textShadow: "2px 2px 0 black",
            textAlign: "center",
            lineHeight: 1.3,
          }}
        >
          {line2}
        </div>
      )}
    </div>
  );
};

import { useCurrentFrame, useVideoConfig } from "remotion";

interface ProgressBarProps {
  durationInSeconds: number;
}

/**
 * Thin progress bar at the bottom of the video.
 * White bar that fills left-to-right over the video duration.
 */
export const ProgressBar: React.FC<ProgressBarProps> = ({
  durationInSeconds,
}) => {
  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const totalFrames = Math.ceil(durationInSeconds * fps);
  const progress = Math.min(frame / totalFrames, 1);

  return (
    <div
      style={{
        position: "absolute",
        bottom: 0,
        left: 0,
        width: 1080,
        height: 4,
        backgroundColor: "rgba(255, 255, 255, 0.2)",
      }}
    >
      <div
        style={{
          width: 1080 * progress,
          height: 4,
          backgroundColor: "white",
        }}
      />
    </div>
  );
};

import { OffthreadVideo, useCurrentFrame, useVideoConfig, interpolate } from "remotion";
import type { Crop, CropKeyframe } from "../types";

interface VideoFrameProps {
  clipSrc: string;
  sourceWidth: number;
  sourceHeight: number;
  crop: Crop;
  cropKeyframes?: CropKeyframe[];
}

/**
 * Renders the reframed video using OffthreadVideo + CSS transforms.
 *
 * Supports two modes:
 * 1. Static crop: uses crop.x directly (no keyframes)
 * 2. Animated pan: interpolates crop.x from keyframes (cursor tracking)
 *
 * Strategy: The source video is scaled and positioned so that only the
 * cropped region is visible within the 1080x1920 container. Uses
 * overflow:hidden to clip, and CSS transform to position the crop area.
 */
export const VideoFrame: React.FC<VideoFrameProps> = ({
  clipSrc,
  sourceWidth,
  sourceHeight,
  crop,
  cropKeyframes,
}) => {
  if (!clipSrc) return null;

  const frame = useCurrentFrame();
  const { fps } = useVideoConfig();
  const currentTime = frame / fps;

  // Scale factor: how much to scale source so crop.w fills 1080px
  const outputWidth = 1080;
  const outputHeight = 1920;
  const scale = outputWidth / crop.w;

  // Determine current crop X (animated or static)
  let cropX = crop.x;
  if (cropKeyframes && cropKeyframes.length >= 2) {
    const times = cropKeyframes.map((k) => k.t);
    const xPositions = cropKeyframes.map((k) => k.x);
    cropX = interpolate(currentTime, times, xPositions, {
      extrapolateLeft: "clamp",
      extrapolateRight: "clamp",
    });
  }

  // Scaled source dimensions
  const scaledWidth = sourceWidth * scale;
  const scaledHeight = sourceHeight * scale;

  // Offset to position the crop region at (0,0)
  const translateX = -cropX * scale;
  const translateY = -crop.y * scale;

  // Auto-detect framed layout: video doesn't fill full output height.
  // Position in content zone (below 200px hook text area) for screen recordings.
  const hookZone = 200;
  const topOffset = scaledHeight < outputHeight ? hookZone : 0;

  return (
    <div
      style={{
        position: "absolute",
        top: topOffset,
        left: 0,
        width: outputWidth,
        height: scaledHeight < outputHeight ? scaledHeight : outputHeight,
        overflow: "hidden",
      }}
    >
      <OffthreadVideo
        src={clipSrc}
        style={{
          position: "absolute",
          width: scaledWidth,
          height: scaledHeight,
          transform: `translate(${translateX}px, ${translateY}px)`,
        }}
      />
    </div>
  );
};

import { useMemo } from "react";
import { createTikTokStyleCaptions } from "@remotion/captions";
import type { Caption } from "../types";

/**
 * Converts word-level captions into TikTok-style pages using
 * @remotion/captions createTikTokStyleCaptions().
 *
 * Each page groups N words together based on combineTokensWithinMilliseconds.
 * Returns pages with: text, startMs, durationMs, tokens[{text, fromMs, toMs}]
 */
export const useCaptionPages = (
  captions: Caption[],
  combineMs: number = 800
) => {
  return useMemo(() => {
    if (!captions || captions.length === 0) return [];

    // Convert our Caption format to Remotion's expected format
    const remotionCaptions = captions.map((c) => ({
      text: c.text,
      startMs: c.startMs,
      endMs: c.endMs,
      timestampMs: null as number | null,
      confidence: null as number | null,
    }));

    const { pages } = createTikTokStyleCaptions({
      captions: remotionCaptions,
      combineTokensWithinMilliseconds: combineMs,
    });

    return pages;
  }, [captions, combineMs]);
};

import { registerRoot } from "remotion";
import { Root } from "./Root";

registerRoot(Root);

import { Composition } from "remotion";
import { ShortVideo } from "./ShortVideo";
import { ShortVideoPropsSchema } from "./types";

export const Root: React.FC = () => {
  return (
    <>
      <Composition
        id="ShortVideo"
        component={ShortVideo}
        width={1080}
        height={1920}
        fps={30}
        durationInFrames={30 * 30}
        schema={ShortVideoPropsSchema}
        defaultProps={{
          clipSrc: "",
          sourceWidth: 1920,
          sourceHeight: 1080,
          crop: { x: 0, y: 0, w: 607, h: 1080 },
          cropKeyframes: [],
          captions: [],
          captionStyle: "bold" as const,
          hookLine1: "",
          hookLine2: "",
          showProgressBar: true,
          durationInSeconds: 30,
        }}
        calculateMetadata={({ props }) => {
          return {
            durationInFrames: Math.ceil(props.durationInSeconds * 30),
            fps: 30,
            width: 1080,
            height: 1920,
          };
        }}
      />
    </>
  );
};

import { AbsoluteFill } from "remotion";
import { VideoFrame } from "./components/VideoFrame";
import { Captions } from "./components/Captions";
import { HookOverlay } from "./components/HookOverlay";
import { ProgressBar } from "./components/ProgressBar";
import { fontFaceCSS } from "./styles/fonts";
import type { ShortVideoProps } from "./types";

export const ShortVideo: React.FC<ShortVideoProps> = ({
  clipSrc,
  sourceWidth,
  sourceHeight,
  crop,
  cropKeyframes,
  captions,
  captionStyle,
  hookLine1,
  hookLine2,
  showProgressBar,
  durationInSeconds,
}) => {
  return (
    <AbsoluteFill style={{ backgroundColor: "black" }}>
      {/* Load custom fonts for captions and hook overlay */}
      <style dangerouslySetInnerHTML={{ __html: fontFaceCSS }} />

      {/* Reframed video (cropped and scaled to fill 1080x1920) */}
      <VideoFrame
        clipSrc={clipSrc}
        sourceWidth={sourceWidth}
        sourceHeight={sourceHeight}
        crop={crop}
        cropKeyframes={cropKeyframes}
      />

      {/* Animated captions */}
      <Captions captions={captions} style={captionStyle} />

      {/* Hook text overlay (first 3.5 seconds) */}
      {(hookLine1 || hookLine2) && (
        <HookOverlay line1={hookLine1 ?? ""} line2={hookLine2 ?? ""} />
      )}

      {/* Progress bar at bottom */}
      {showProgressBar && (
        <ProgressBar durationInSeconds={durationInSeconds} />
      )}
    </AbsoluteFill>
  );
};

import { staticFile } from "remotion";

/**
 * Font declarations for caption styles.
 *
 * Fonts are loaded from public/fonts/ directory:
 * - Montserrat Bold (Bold style) — Google Fonts, OFL
 * - Bangers (Bounce style) — Google Fonts, OFL
 * - Inter Bold (Clean style) — Google Fonts, OFL
 *
 * For preview mode, fonts are loaded via @font-face in the browser.
 * For rendering, Remotion handles font loading automatically.
 */
export const FONTS = {
  montserrat: {
    family: "Montserrat",
    src: staticFile("fonts/Montserrat-Bold.ttf"),
    weight: "800",
  },
  bangers: {
    family: "Bangers",
    src: staticFile("fonts/Bangers-Regular.ttf"),
    weight: "400",
  },
  inter: {
    family: "Inter",
    src: staticFile("fonts/Inter-Bold.ttf"),
    weight: "700",
  },
} as const;

/**
 * CSS @font-face declarations for all caption fonts.
 * Inject into document head for preview mode.
 */
export const fontFaceCSS = Object.values(FONTS)
  .map(
    (f) => `
@font-face {
  font-family: '${f.family}';
  src: url('${f.src}') format('truetype');
  font-weight: ${f.weight};
  font-display: block;
}
`
  )
  .join("\n");

/**
 * Color palettes and style constants for each caption preset.
 */

export const BOLD_THEME = {
  textColor: "#FFFFFF",
  activeColor: "#FFD700", // Yellow highlight on active word
  shadowColor: "#000000",
  backgroundColor: "transparent",
};

export const BOUNCE_THEME = {
  textColor: "#FFFFFF",
  shadowColor: "#000000",
  backgroundColor: "transparent",
  // Rotating bright colors per page
  rotatingColors: [
    "#00FFFF", // Cyan
    "#FF00FF", // Magenta
    "#00FF00", // Lime
    "#FFFF00", // Yellow
    "#FF6600", // Orange
    "#FF0066", // Hot pink
  ],
};

export const CLEAN_THEME = {
  textColor: "#FFFFFF",
  activeColor: "#E0E0E0", // Subtle lighter white for active word
  shadowColor: "rgba(0, 0, 0, 0.6)",
  backgroundColor: "transparent",
};

import { z } from "zod";

export const CaptionSchema = z.object({
  text: z.string(),
  startMs: z.number(),
  endMs: z.number(),
});

export type Caption = z.infer<typeof CaptionSchema>;

export const CropSchema = z.object({
  x: z.number(),
  y: z.number(),
  w: z.number(),
  h: z.number(),
});

export type Crop = z.infer<typeof CropSchema>;

export const CropKeyframeSchema = z.object({
  t: z.number(),
  x: z.number(),
});

export type CropKeyframe = z.infer<typeof CropKeyframeSchema>;

export const CaptionStyle = z.enum(["bold", "bounce", "clean"]);
export type CaptionStyleType = z.infer<typeof CaptionStyle>;

export const ShortVideoPropsSchema = z.object({
  clipSrc: z.string(),
  sourceWidth: z.number(),
  sourceHeight: z.number(),
  crop: CropSchema,
  cropKeyframes: z.array(CropKeyframeSchema).optional().default([]),
  captions: z.array(CaptionSchema),
  captionStyle: CaptionStyle,
  hookLine1: z.string().optional().default(""),
  hookLine2: z.string().optional().default(""),
  showProgressBar: z.boolean().optional().default(true),
  durationInSeconds: z.number(),
});

export type ShortVideoProps = z.infer<typeof ShortVideoPropsSchema>;

{
  "compilerOptions": {
    "target": "ES2022",
    "module": "ES2022",
    "moduleResolution": "bundler",
    "jsx": "react-jsx",
    "strict": true,
    "esModuleInterop": true,
    "skipLibCheck": true,
    "forceConsistentCasingInFileNames": true,
    "resolveJsonModule": true,
    "outDir": "dist",
    "baseUrl": ".",
    "paths": {
      "@/*": ["src/*"]
    }
  },
  "include": ["src/**/*"],
  "exclude": ["node_modules", "dist"]
}

faster-whisper>=1.0.0
mediapipe>=0.10.0
numpy>=1.24.0
opencv-python>=4.8.0

#!/usr/bin/env python3
"""Compute reframe coordinates for vertical (9:16) video from landscape clips.

Supports three reframing strategies based on content type:
- talking-head: Face-tracked center crop (follows the speaker's face)
- screen: Cursor-tracked pan with moderate zoom (follows mouse cursor)
- podcast: Center crop or dominant-speaker tracking

Usage:
    python3 compute_reframe.py --clips-dir /tmp/claude-shorts/clips/ \
        --content-type screen --output reframe.json

Output (reframe.json):
{
    "clips": {
        "clip_01.mp4": {
            "strategy": "cursor-track",
            "source_resolution": "3840x2160",
            "crop": {"x": 800, "y": 0, "w": 2112, "h": 2160},
            "crop_keyframes": [
                {"t": 0.0, "x": 800},
                {"t": 1.0, "x": 950}
            ],
            "output_resolution": "1080x1920"
        }
    }
}
"""
import argparse
import glob
import json
import os
import subprocess
import sys
import time


def get_video_info(path):
    """Get video width, height, fps, and duration."""
    result = subprocess.run(
        ["ffprobe", "-v", "quiet", "-print_format", "json",
         "-show_streams", "-show_format", "-select_streams", "v:0", path],
        capture_output=True, text=True
    )
    data = json.loads(result.stdout)
    stream = data["streams"][0]
    fmt = data.get("format", {})
    w = int(stream["width"])
    h = int(stream["height"])
    duration = float(fmt.get("duration", stream.get("duration", 0)))
    fps_str = stream.get("r_frame_rate", "30/1")
    num, den = fps_str.split("/")
    fps = round(float(num) / float(den))
    return w, h, fps, duration


def compute_face_track(clip_path, src_w, src_h, num_samples=5):
    """Track face positions across sampled frames using MediaPipe."""
    import cv2
    import mediapipe as mp
    import tempfile

    _, _, _, duration = get_video_info(clip_path)

    mp_face = mp.solutions.face_detection
    detector = mp_face.FaceDetection(model_selection=1, min_detection_confidence=0.5)

    face_positions = []
    timestamps = [duration * (i + 0.5) / num_samples for i in range(num_samples)]

    tmpdir = tempfile.mkdtemp()

    for ts in timestamps:
        frame_path = os.path.join(tmpdir, f"f_{ts:.2f}.jpg")
        subprocess.run(
            ["ffmpeg", "-y", "-ss", str(ts), "-i", clip_path,
             "-vframes", "1", "-q:v", "2", frame_path],
            capture_output=True
        )

        if not os.path.exists(frame_path):
            continue

        img = cv2.imread(frame_path)
        if img is None:
            continue

        rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
        results = detector.process(rgb)

        if results.detections:
            best = max(results.detections,
                       key=lambda d: d.location_data.relative_bounding_box.width *
                                     d.location_data.relative_bounding_box.height)
            bbox = best.location_data.relative_bounding_box
            cx = bbox.xmin + bbox.width / 2
            cy = bbox.ymin + bbox.height / 2
            face_positions.append({
                "t": round(ts, 2),
                "x_center": round(cx, 4),
                "y_center": round(cy, 4),
            })

        os.unlink(frame_path)

    detector.close()

    import shutil
    shutil.rmtree(tmpdir, ignore_errors=True)

    return face_positions


def detect_cursor_positions(clip_path, src_w, src_h, sample_interval=0.5):
    """Detect mouse cursor positions via frame differencing.

    Compares consecutive frames to find small changed regions (cursor movement).
    Returns list of {t, x_norm, y_norm} positions.
    """
    import cv2
    import numpy as np
    import tempfile

    _, _, _, duration = get_video_info(clip_path)

    if duration < sample_interval * 2:
        return []

    tmpdir = tempfile.mkdtemp()

    # Extract all sample frames in a single FFmpeg call (much faster than per-frame seeks)
    fps_rate = 1.0 / sample_interval
    subprocess.run(
        ["ffmpeg", "-y", "-i", clip_path,
         "-vf", f"fps={fps_rate}", "-q:v", "2",
         os.path.join(tmpdir, "f_%04d.jpg")],
        capture_output=True
    )

    # Read extracted frames in order (FFmpeg numbers from 0001)
    import shutil
    frame_files = sorted(glob.glob(os.path.join(tmpdir, "f_*.jpg")))
    frames = []
    for i, frame_path in enumerate(frame_files):
        ts = i * sample_interval
        if ts >= duration:
            break
        img = cv2.imread(frame_path, cv2.IMREAD_GRAYSCALE)
        frames.append((ts, img))

    shutil.rmtree(tmpdir, ignore_errors=True)

    if len(frames) < 2:
        return []

    cursor_positions = []
    last_known_x = 0.5  # default to center

    for i in range(1, len(frames)):
        ts, curr = frames[i]
        _, prev = frames[i - 1]

        if curr is None or prev is None:
            cursor_positions.append({"t": round(ts, 2), "x_norm": last_known_x})
            continue

        # Compute absolute difference
        diff = cv2.absdiff(prev, curr)

        # Threshold to find changed pixels
        _, thresh = cv2.threshold(diff, 25, 255, cv2.THRESH_BINARY)

        # Find contours of changed regions
        contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

        # Filter for cursor-sized changes (small moving elements)
        # Cursor is typically 10-60px in each dimension
        cursor_candidates = []
        for c in contours:
            area = cv2.contourArea(c)
            x, y, w, h = cv2.boundingRect(c)
            # Cursor-sized: 50-5000 px² area, aspect ratio not too extreme
            if 50 < area < 5000 and 0.2 < (w / max(h, 1)) < 5.0:
                cx = (x + w / 2) / src_w
                cy = (y + h / 2) / src_h
                cursor_candidates.append({
                    "x_norm": cx,
                    "y_norm": cy,
                    "area": area,
                    "dist_from_last": abs(cx - last_known_x),
                })

        if cursor_candidates:
            # Pick the candidate closest to last known position (smooth tracking)
            best = min(cursor_candidates, key=lambda c: c["dist_from_last"])
            last_known_x = best["x_norm"]
            cursor_positions.append({"t": round(ts, 2), "x_norm": round(best["x_norm"], 4)})
        else:
            # No cursor-sized change detected; could be:
            # - cursor not moving (use last known)
            # - large scroll event (ignore, keep position)
            cursor_positions.append({"t": round(ts, 2), "x_norm": round(last_known_x, 4)})

    # Add initial position (use first detected or center)
    if cursor_positions:
        cursor_positions.insert(0, {"t": 0.0, "x_norm": cursor_positions[0]["x_norm"]})

    return cursor_positions


def smooth_positions(positions, window=5):
    """Apply moving average smoothing to cursor positions."""
    if len(positions) < window:
        return positions

    import numpy as np
    x_vals = np.array([p["x_norm"] for p in positions])

    # Pad edges for convolution
    kernel = np.ones(window) / window
    smoothed = np.convolve(x_vals, kernel, mode="same")

    # Fix edges (use original values for first/last few)
    half = window // 2
    smoothed[:half] = x_vals[:half]
    smoothed[-half:] = x_vals[-half:]

    result = []
    for i, p in enumerate(positions):
        result.append({"t": p["t"], "x_norm": round(float(smoothed[i]), 4)})

    return result


def compute_crop_screen_with_cursor(src_w, src_h, cursor_positions, zoom=0.55):
    """Compute cursor-tracked crop for screen recordings.

    Args:
        src_w, src_h: Source video dimensions
        cursor_positions: List of {t, x_norm} from detect_cursor_positions
        zoom: Fraction of source width to show (0.55 = 55% of screen visible)

    Returns:
        (crop, crop_keyframes) where:
        - crop: static fallback {x, y, w, h}
        - crop_keyframes: list of {t, x} for animated panning
    """
    crop_h = src_h
    crop_w = int(src_w * zoom)
    crop_w = min(crop_w, src_w)

    # Compute crop_x for each cursor position
    keyframes = []
    for pos in cursor_positions:
        cursor_x_px = pos["x_norm"] * src_w
        crop_x = int(cursor_x_px - crop_w / 2)
        crop_x = max(0, min(crop_x, src_w - crop_w))
        keyframes.append({"t": round(pos["t"], 2), "x": crop_x})

    # Deduplicate: remove keyframes where x hasn't changed much
    if len(keyframes) > 2:
        filtered = [keyframes[0]]
        for i in range(1, len(keyframes) - 1):
            prev_x = filtered[-1]["x"]
            curr_x = keyframes[i]["x"]
            # Keep keyframe if position changed by more than 2% of crop width
            if abs(curr_x - prev_x) > crop_w * 0.02:
                filtered.append(keyframes[i])
        filtered.append(keyframes[-1])
        keyframes = filtered

    # Static fallback: average position
    if keyframes:
        avg_x = int(sum(k["x"] for k in keyframes) / len(keyframes))
    else:
        avg_x = (src_w - crop_w) // 2

    avg_x = max(0, min(avg_x, src_w - crop_w))

    crop = {"x": avg_x, "y": 0, "w": crop_w, "h": crop_h}

    return crop, keyframes


def compute_crop_screen_static(src_w, src_h, zoom=0.55):
    """Static center crop for screen recordings (no cursor tracking)."""
    crop_h = src_h
    crop_w = int(src_w * zoom)
    crop_w = min(crop_w, src_w)

    center_x = int(src_w * 0.50)
    crop_x = center_x - crop_w // 2
    crop_x = max(0, min(crop_x, src_w - crop_w))

    return {"x": crop_x, "y": 0, "w": crop_w, "h": crop_h}


def compute_crop_face_track(src_w, src_h, face_positions):
    """Compute 9:16 crop centered on average face position."""
    crop_h = src_h
    crop_w = int(crop_h * 9 / 16)
    crop_w = min(crop_w, src_w)

    if face_positions:
        avg_cx = sum(fp["x_center"] for fp in face_positions) / len(face_positions)
        center_x_px = int(avg_cx * src_w)
    else:
        center_x_px = src_w // 2

    crop_x = center_x_px - crop_w // 2
    crop_x = max(0, min(crop_x, src_w - crop_w))

    return {"x": crop_x, "y": 0, "w": crop_w, "h": crop_h}


def compute_crop_center(src_w, src_h):
    """Simple center crop to 9:16."""
    crop_h = src_h
    crop_w = int(crop_h * 9 / 16)
    crop_w = min(crop_w, src_w)
    crop_x = (src_w - crop_w) // 2

    return {"x": crop_x, "y": 0, "w": crop_w, "h": crop_h}


def main():
    parser = argparse.ArgumentParser(description="Compute reframe coordinates")
    parser.add_argument("--clips-dir", required=True, help="Directory with clip files")
    parser.add_argument("--content-type", required=True,
                        choices=["talking-head", "screen", "podcast"],
                        help="Content type for reframing strategy")
    parser.add_argument("--output", required=True, help="Output JSON file")
    parser.add_argument("--zoom", type=float, default=0.55,
                        help="Screen zoom level: fraction of source width to show (default: 0.55)")
    parser.add_argument("--no-cursor-track", action="store_true",
                        help="Disable cursor tracking for screen content")

    args = parser.parse_args()

    if not os.path.isdir(args.clips_dir):
        print(json.dumps({"error": f"Clips directory not found: {args.clips_dir}"}))
        sys.exit(1)

    start = time.time()

    clips = sorted(glob.glob(os.path.join(args.clips_dir, "clip_*.mp4")))
    if not clips:
        print(json.dumps({"error": f"No clip_*.mp4 files found in {args.clips_dir}"}))
        sys.exit(1)

    results = {}

    for clip_path in clips:
        clip_name = os.path.basename(clip_path)
        src_w, src_h, fps, duration = get_video_info(clip_path)

        if args.content_type == "talking-head":
            face_positions = compute_face_track(clip_path, src_w, src_h)
            crop = compute_crop_face_track(src_w, src_h, face_positions)
            strategy = "face-track"
            crop_keyframes = []
        elif args.content_type == "screen":
            if args.no_cursor_track:
                crop = compute_crop_screen_static(src_w, src_h, zoom=args.zoom)
                crop_keyframes = []
                strategy = "framed"
            else:
                # Detect cursor positions
                cursor_positions = detect_cursor_positions(
                    clip_path, src_w, src_h, sample_interval=0.5
                )
                # Smooth the trajectory
                cursor_positions = smooth_positions(cursor_positions, window=5)

                if len(cursor_positions) >= 2:
                    crop, crop_keyframes = compute_crop_screen_with_cursor(
                        src_w, src_h, cursor_positions, zoom=args.zoom
                    )
                    strategy = "cursor-track"
                else:
                    crop = compute_crop_screen_static(src_w, src_h, zoom=args.zoom)
                    crop_keyframes = []
                    strategy = "framed"
            face_positions = []
        else:  # podcast
            face_positions = compute_face_track(clip_path, src_w, src_h)
            crop = compute_crop_face_track(src_w, src_h, face_positions)
            strategy = "face-track"
            crop_keyframes = []

        entry = {
            "strategy": strategy,
            "source_resolution": f"{src_w}x{src_h}",
            "crop": crop,
            "output_resolution": "1080x1920",
            "face_positions": face_positions if args.content_type != "screen" else [],
            "duration": round(duration, 2),
        }

        if crop_keyframes:
            entry["crop_keyframes"] = crop_keyframes

        results[clip_name] = entry

    elapsed = time.time() - start

    output = {
        "content_type": args.content_type,
        "clip_count": len(results),
        "computation_time_sec": round(elapsed, 2),
        "clips": results,
    }

    os.makedirs(os.path.dirname(os.path.abspath(args.output)), exist_ok=True)
    with open(args.output, "w") as f:
        json.dump(output, f, indent=2)

    # Print summary
    summary = {
        "action": "compute_reframe",
        "content_type": args.content_type,
        "clips_processed": len(results),
        "computation_time_sec": round(elapsed, 2),
    }
    for name, data in results.items():
        s = {
            "strategy": data["strategy"],
            "crop": data["crop"],
        }
        if "crop_keyframes" in data:
            s["keyframe_count"] = len(data["crop_keyframes"])
        if data.get("face_positions"):
            s["faces_detected"] = len(data["face_positions"])
        summary[name] = s
    print(json.dumps(summary, indent=2))


if __name__ == "__main__":
    main()