ReelForge

Auto-caption & auto-edit raw clips into vertical 9:16 reels — faster-whisper + ffmpeg, driven by one Claude Code command.

Free, local, no watermark, no subscription, no per-clip credits. A self-hostable, open-source alternative to Opus Clip / SubMagic / CapCut auto-captions — for content creators, freelancers, and indie hackers.

Status

Last reviewed: September 2026 · release v2026.09

What is ReelForge?

ReelForge is a free, open-source CLI that turns one raw talking-head take into a finished, captioned, hook-carded vertical reel — with no manual editing. It transcribes locally with faster-whisper (no API key, no cloud), burns word-synced captions, drops an on-screen hook card, crops to 1080×1920, and exports — all from a single command or a Claude Code /reel skill.

Unlike "viral clip" tools that need a long podcast and try to guess the good moments, ReelForge assumes you filmed one intentional take and just want it shipped. Record → run → publish.

The differentiator: single-take input, a real hook card, baked-in branding, zero config, and Claude Code native. Most open-source tools do none of these.

How do I auto-caption a video? (Quickstart)

With Claude Code — one command

npm install -g @anthropic-ai/claude-code        # 1. install Claude Code
git clone https://github.com/waseemnasir2k26/reelforge.git && cd reelforge   # 2. clone
claude                                           # 3. open Claude Code here

Then, inside Claude Code:

/reel myclip.mp4 "THIS REEL EDITED ITSELF"

Claude installs the deps, runs the pipeline, and hands you reel.mp4. Done.

Or run it raw (no Claude Code)

pip install -r requirements.txt        # faster-whisper
# ffmpeg must be installed and on PATH — https://ffmpeg.org/download.html

python auto_reel.py myclip.mp4 \
  --hook "THIS REEL EDITED ITSELF" \
  --sub  "FULLY AI / ZERO HUMAN EDITOR" \
  --out  reel.mp4

How does it work?

raw take ──▶ faster-whisper ──▶ word-synced ──▶ hook card ──▶ scale/crop ──▶ ffmpeg ──▶ reel.mp4
 (1 take)   (local, $0, no key)   ASS captions   (drawtext)    1080×1920    NVENC│x264

1. Transcribe — faster-whisper returns word-level timestamps, locally, for free.
2. Caption — builds a burned-in ASS subtitle track, snapped to each word.
3. Hook card — composites your hook + subline over the first 2.5s.
4. Reframe — scales and center-crops any input to vertical 1080×1920.
5. Export — one ffmpeg pass via h264_nvenc (GPU), auto-falling back to libx264 (CPU).

Options

FAQ

Is this a free Opus Clip / SubMagic alternative?
Yes. ReelForge runs entirely on your machine — no watermark, no subscription, no per-clip credits, no upload limits.

Does it need an API key or internet?
No. faster-whisper runs locally. The only requirements are Python and ffmpeg.

What ffmpeg command burns subtitles into a vertical video?
ReelForge does it for you, but under the hood it scales to 1080×1920, center-crops, and applies a subtitles= filter plus a drawtext hook card in a single pass.

Can Claude Code edit my videos and add captions automatically?
Yes — that's what the bundled /reel command and .claude/skills/reel/SKILL.md do.

Does it work without a GPU?
Yes. It tries NVENC first and automatically falls back to CPU libx264.

What platforms is the output for?
TikTok, Instagram Reels, YouTube Shorts — any 1080×1920 9:16 destination.

Requirements

• Python 3.9+
• ffmpeg on PATH — download
• faster-whisper (installed via requirements.txt)
• GPU optional (NVENC); CPU works via libx264 fallback

Roadmap

• Auto silence-trim (cut dead air via silencedetect)
• Karaoke / word-highlight caption style
• Brand preset file (logo, handle, colors)
• Batch mode (folder → folder)
• Per-platform caption packs on export

Contributions welcome — see CONTRIBUTING.md.

Built by SkynetLabs

ReelForge is the free, do-it-yourself version. SkynetLabs builds the full pipeline for brands — tuned to your voice and platforms, auto-publishing daily, so your team stops editing by hand.

👉 www.skynetjoe.com · DM "SKILL"