6 open-source tools that subtitle your video with Whisper

Speech-to-text APIs bill by the minute, and web subtitle tools upload your footage to someone else's server. These six open-source Whisper tools transcribe on your own machine and hand you SRT files (the standard subtitle format) any editor can load, for nothing.

Every pick is open source, pushed a commit in the last 90 days (checked 2026-11-11), does what its docs claim, and runs offline once its model has downloaded. Ordered as the actual subtitle workflow: transcribe, fix, style, burn.

The 6 tools at a glance #

1. Vibe: transcribe a video in one drag #

Vibe is a desktop app (macOS, Windows, Linux) that transcribes audio and video entirely offline — drop a file, a folder, or a YouTube URL, and export SRT, VTT, TXT, HTML, PDF or DOCX.

What: Local Whisper-based transcription with batch files, speaker diarization (who spoke when) and a stable-timestamps mode tuned for subtitle-grade timing.

Why: No upload, no per-minute meter — the model downloads once and every transcript after that is free and private.

Use when: You want subtitle files without touching a terminal.

Install: Download the installer for your OS from thewh1teagle.github.io/vibe.

Catch: The stable-timestamps mode trades speed for accuracy and runs slower than normal transcription.

2. WhisperX: word-accurate timing from the terminal #

WhisperX transcribes up to 70× realtime with batched inference, then aligns every word to the audio with wav2vec2 forced alignment — so the SRT it writes is timed per word, not per guessed segment. Speaker diarization labels who said what.

What: A Python CLI that outputs SRT/VTT with word-level timestamps and optional speaker labels.

Why: Whisper's native timestamps drift by seconds; forced alignment pulls them back to the word.

Use when: Long videos, multiple speakers, or a transcription step you want to script and re-run.

Install:

pip install whisperx

Catch: Speaker diarization needs a free Hugging Face token passed as --hf_token plus accepting the pyannote model's user agreement.

3. whisper.cpp: the engine, everywhere #

whisper.cpp is Whisper ported to plain C/C++ — no Python, no CUDA required. Its whisper-cli tool transcribes on CPU, Apple Silicon, Windows, Docker, even a Raspberry Pi, and writes files with -osrt.

What: The engine behind many transcription apps, run straight from the command line.

Why: Runs where the Python stack can't — a NAS, a Pi, an old laptop — and starts in seconds.

Use when: You want transcription embedded in your own tool or on modest hardware.

Install:

git clone https://github.com/ggml-org/whisper.cpp && cd whisper.cpp && cmake -B build && cmake --build build -j --config Release

Catch: whisper-cli takes 16-bit WAV only — convert first with ffmpeg -i input.mp3 -ar 16000 -ac 1 -c:a pcm_s16le output.wav.

4. Subtitle Edit: fix the timing Whisper got wrong #

Subtitle Edit is a full subtitle editor with Whisper speech-to-text built in: Fix Common Errors, Visual Sync, duration limits, and batch conversion between SRT, VTT and ASS.

What: The cleanup pass: shift, split, merge, sync and convert subtitle files.

Why: Machine transcripts are rarely publish-ready; this is where they become it.

Use when: After any tool above, before the video ships.

Install: Download the signed installer from Releases — the macOS .dmg is notarized, Linux ships as a Flatpak.

Catch: Windows needs Windows 10 22H2 or newer — older builds can fail to start with a .NET runtime error.

5. Aegisub: style captions like a typesetter #

Aegisub is the classic subtitle editor for styling: colors, fonts, positioning, motion and karaoke effects through ASS override tags, drawn against the video frame.

What: An editor built around ASS, the subtitle format that can actually carry styling.

Why: SRT has no styles; when captions need brand colors and placement, ASS is the format and Aegisub is where it gets written.

Use when: Styled captions for YouTube or reels rather than plain white text.

Install: Get the stable build from aegisub.org or the latest upstream build from its nightly releases.

Catch: No transcription inside — bring the SRT from steps 1–3, and expect a real learning curve.

6. ffmpeg: burn it in, one command #

ffmpeg is the open-source media toolkit; its subtitles filter renders an SRT or ASS file onto the video with libass:

ffmpeg -i video.mp4 -vf "subtitles=captions.srt" out.mp4

What: Captions become pixels, for platforms that ignore subtitle files.

Why: For reels and shorts you need text on the frames; ffmpeg does it without opening an editor.

Use when: Exporting the finished file.

Install: Get a full build from the ffmpeg.org download page; run ffmpeg -filters | grep subtitles to confirm it can burn captions.

Catch: The subtitles filter needs libass. Homebrew's ffmpeg and other minimal builds ship without it, so the burn command fails there.

Which one to try first #

Never opened a terminal → #1. Long video, several speakers → #2. Already have a subtitle file → #4. Plain captions for a reel → #6. If you'd rather transcribe in the cloud on free API credits instead of locally, that's our free AI subtitle generator guide.