Building Capite: a self-hosted AI video caption generator with faster-whisper and FFmpeg

Description and use

Capite is a fully local, subscription‑free caption generator using faster‑whisper and FFmpeg, offering 27 motion presets and 100+ languages.

Problem

Short‑form video creators rely on paid, cloud‑based caption tools that impose subscription fees, minute limits, and data privacy concerns.

Outcome

Capite delivers fully local, subscription‑free caption generation with 27 motion presets, 100+ languages, and sub‑5‑second re‑encoding after edits, eliminating external dependencies and privacy risks.

Workflow

• Clone the Capite repo
• Navigate to the project directory
• Run docker compose up to start services
• Open http://localhost:3000 in a browser
• Upload a video
• Transcription is performed by faster‑whisper (CTranslate2) to get millisecond word timestamps
• Edit the transcript in the Next.js editor (click words, adjust timing)
• Choose a motion preset from 27 available options
• Compile subtitles with pysubs2 into an .ass script
• FFmpeg with libass burns captions into the video at CRF 18 or exports raw subtitle files
• Export final video or subtitle files in MP4, .srt, .vtt, .txt, or .ass

Evidence

Capite transcribes video audio locally, aligns word boundaries to the millisecond, applies one of 27 motion typography presets, and burns it into the video at broadcast quality.

Source

Re‑encoding after an edit takes under 5 seconds – it skips the transcription step entirely.

Source