v0.4.0 — OpenAI API backend + large-v3-turbo

What's New

OpenAI API Transcription Backend

Use OpenAI's transcription API instead of local Whisper for max speed:

trx init --backend openai
trx video.mp4 -b openai

Three API models: gpt-4o-transcribe (best accuracy), gpt-4o-mini-transcribe (cheapest), whisper-1 (legacy with timestamps).

large-v3-turbo Local Model

Near-large accuracy at ~3x speed:

trx init --model large-v3-turbo

Cross-Platform Support

trx init now works on Linux (apt-get + cmake build) and Windows (winget + pre-built binaries). Thanks @isjunrod! (#1, #2)

Changelog & Docs

• trx.crafter.run/changelog
• trx.crafter.run/docs

Other

• OS-specific open command (macOS/Linux/Windows)
• Quoted paths with spaces for ctrl+click support
• Config defaults merge prevents crashes on missing fields
• FFmpeg path collision fix for .wav input files
• SRT timestamps use real audio duration via ffprobe

Install / Upgrade

bun add -g @crafter/[email protected]

Full Changelog: v0.3.2...v0.4.0

v0.2.1

What's New

Features

• Progress bar during Whisper transcription

Fixes

• Spinner hang after transcription — guard callbacks + exit

Changes since v0.2.0

• feat: progress bar during Whisper transcription
• fix: spinner hang after transcription, guard callbacks + exit

Install

bun add -g @crafter/trx

npm package