Cameraman

Open-source screen recorder and video editor for macOS. Swift native. Local-first. Free.

Version: 0.7.1 (beta) · Platform: macOS 13+ (Ventura)

Landing page: the marketing site lives in a separate repo — dPeluChe/cameraman-landing (React 19 + Tailwind 4 + framer-motion, deployed to cameraman.dev via Vercel). Features an animated editor mockup, bento feature grid with honest status badges (shipped / being tuned / improving), MCP agent chat mock, and comparison table. The static docs/index.html previously used for GitHub Pages is superseded but kept for backwards compatibility.

---

What it does

Cameraman captures your screen, camera, and audio as separate tracks, gives you a non-destructive timeline editor on top, and exports to standard video formats. Everything happens locally — nothing leaves your machine.

Recording

• Screen capture via ScreenCaptureKit
• Camera (webcam) overlay
• System audio + microphone (separate tracks)
• Cursor/click telemetry for auto-zoom

Editing

• Multi-track timeline (trim, split, delete, drag-and-drop)
• Per-segment speed control (0.25× – 4×)
• Per-segment camera PiP positioning with borders (circle, rounded rect, capsule)
• Per-segment audio volume (0–300%) and mute
• Overlays: arrows, rectangles, lines, text with timing controls
• Per-clip effects: extensible, non-destructive adjustments (sepia, B&W, brightness/contrast/saturation, vibrance, hue, invert, vignette, blur; audio pitch) targetable to a layer (screen / camera / background) over a time range — applied in preview and export
• Canvas effects: background gradients, blur, padding, corner radius
• Auto-zoom suggestions from cursor telemetry (click + dwell detection)
• Import external videos (with their audio) onto their own timeline rows — magnetic snap, trim, split, PiP positioning, row reorder
• Merge two projects into a new one (B appended after A)
• Empty projects: edit imported clips without recording anything
• Undo/redo with autosave
• Non-destructive — source files are never modified

Export

• Web 1080p (H.264) — optimized for web sharing
• High 1080p (HEVC) — smaller files, better quality
• 4K (HEVC) — 3840×2160, 60fps
• Portrait 1080p (H.264) — vertical format
• Animated GIF — configurable fps, size, loop
• Quality control (smaller file / standard / higher quality) with a live size estimate
• Per-segment camera positions, visual effects and audio are all preserved in export
• Share whole projects between machines as portable .cameramanproject bundles (now interoperable with Cameraman Mobile)

Captions & transcription

• On-device speech-to-text (WhisperKit, Apple Silicon) with a model picker — keeps the spoken language or translates to English
• Generate from Generate Captions… (⇧⌘T) or the inspector's Captions & AI section; review/edit segments, copy, regenerate, export SRT/VTT
• Add as editable subtitles on the timeline that render over the video (and burn into exports), with a Subtitles visibility toggle
• AI Assistant (MCP) (⇧⌘J): a hub that connects your AI assistant (Claude/Codex) via the bundled MCP server so it can edit the project for you — cut silences, add chapters, caption, apply effects, export

Interface

• Native macOS look with a unified design system; Light / Dark / System appearance (Settings → General → Interface)

Automate (MCP server)

• A built-in MCP server (cameraman-mcp, bundled and signed inside the app) exposes Cameraman to AI assistants (Claude Desktop/Code, Codex) over stdio JSON-RPC — 43 tools to create/edit/record projects, add and edit clips/tracks/overlays/effects, set the canvas, transcribe, and export (with job polling). Register it from Settings → Developer. See MCPServer/README.md.

---

Install (beta)

Pre-built .dmg releases are published to GitHub Releases (when available).

The beta .dmg is not signed with a Developer ID nor notarized — testers will hit a Gatekeeper warning on first launch:

• macOS 14 and earlier: right-click the app in /Applications → Open → confirm.
• macOS 15 (Sequoia) and later: double-click → System Settings → Privacy & Security → Open Anyway.
• macOS 26 (Tahoe): the dialogs above don't bypass quarantine on Tahoe; run once in Terminal:

  xattr -dr com.apple.quarantine /Applications/CameramanApp.app

After the first authorized open, the app launches normally.

---

Build from source

Requirements: macOS 13+, Xcode 15+, Swift 5.9+.

git clone https://github.com/dPeluChe/cameraman.git
cd cameraman
open CameramanApp/CameramanApp.xcodeproj
# Scheme: CameramanApp → My Mac → Cmd+R

For full setup, architecture and patterns see docs/DEV_ONBOARDING.md.

Beta packaging (.dmg)

One command produces a universal (arm64 + x86_64) .dmg ready to ship:

make release

This runs build → verify → dmg in one pass:

1. Compiles a universal Release build
2. Validates that the binary contains both architectures (aborts otherwise)
3. Packages dist/Cameraman-beta-X.Y.Z.B.dmg with the branded background and an Applications drop link

Command	When to use
make release	Default — full beta pipeline (build + verify + dmg)
make build	Universal Release build only (no DMG)
make build-arm	Native arm64-only build for fast local iteration
make verify	Confirm an existing build is universal
make dmg	Re-package the existing build (e.g. after editing the DMG background)
make clean	Remove dist/ and CameramanApp/build/
make help	List all targets

Requirement: brew install create-dmg.

To remove the Gatekeeper warning entirely the app would need an Apple Developer Program subscription, a Developer ID Application certificate, and notarization via xcrun notarytool.

---

Architecture (at a glance)

cameraman/                         # this repo — the macOS app + engine
├── App/                           # SwiftUI app (CameramanApp scheme)
├── CameramanApp/                  # Xcode project, entitlements, assets
├── EngineKit/                     # Pure Swift package — engine code (UI-free)
│   ├── Capture/                   # Recording, permissions, telemetry
│   ├── Editor/                    # Non-destructive editing model + overlays
│   ├── Preview/                   # Playback with edits applied
│   ├── Export/                    # Async rendering pipeline + presets
│   ├── Zoom/                      # Auto-zoom from cursor telemetry
│   ├── Transcription/             # Offline STT (WhisperKit)
│   └── Shared/Models/Store/       # Cross-cutting code + persistence
├── MCPServer/                     # MCP server (cameraman-mcp, 37 tools)
├── docs/                          # CHANGELOG, PRD, TECH_SPEC, DEV_ONBOARDING, TASK_*
│   ├── index.html                 # Legacy static landing (superseded by cameraman-landing)
│   └── branding/                  # App icon, wordmark, DMG background
├── scripts/build-dmg.sh
└── Makefile

cameraman-landing/                 # separate repo — cameraman.dev
└── (React 19 + Tailwind 4 + framer-motion, deployed via Vercel)

Key design choices:

• Non-destructive editing — all edits live in project.json; source media is read-only.
• Multi-track timeline — typed tracks (primary/video/audio) holding universal clips (recording, image, video, audio, color).
• Engine/UI separation — EngineKit exposes a stable API; the SwiftUI layer is replaceable.
• Actor model — CaptureEngine, CameraEngine, PreviewEngine, ThumbnailCache use Swift actors for thread-safe state.
• Job-based processing — export, transcription and proxy generation run as async background jobs.

For details see docs/DEV_ONBOARDING.md and docs/TECH_SPEC.md.

---

Contributing

We welcome contributions! Bug reports, feature requests, docs improvements, code — all are appreciated. Read CONTRIBUTING.md to get set up and learn the conventions. The current backlog lives in docs/TASK_TODO.md.

Quick tips:

• Pick a fix/, feat/, perf/, refactor/, docs/ or chore/ branch prefix.
• Keep PRs focused — one logical unit of work each.
• Engine code is UI-free. Don't add SwiftUI / AppKit dependencies inside EngineKit/.
• Files cap at ~400–500 LOC. Split with extensions when growing.
• Write the why, not the what in comments and commit bodies.

---

Status

Beta. Core recording, editing, and export workflows are functional. Overlay system, auto-zoom and per-segment editing are working. Actively developed.

See docs/CHANGELOG.md for version history. See docs/TASK_TODO.md for planned work.

---

License

MIT — © 2026 Antonio Martinez Quintero (dPeluChe).

---

Built by

dPeluChe — independent developer building tools at dpeluche.dev.