SubX-CLI

English | 中文

AI-powered CLI for automated subtitle matching, renaming, format conversion, and timeline correction.

What SubX Does

Point SubX at a media folder and it uses AI to figure out which subtitle files belong to which videos, even when the filenames are in different languages and share no common naming pattern.

Before:
media/
├── movies/
│   └── The.Matrix.1080p.mkv
└── subtitles/
    ├── Matrix_EN_Sub.srt
    └── 駭客任務_中文字幕.srt

After running: subx-cli match --copy media/
media/
├── movies/
│   ├── The.Matrix.1080p.mkv
│   ├── The.Matrix.1080p.srt         ← AI matched Matrix_EN_Sub.srt
│   └── The.Matrix.1080p.zh.srt      ← AI matched 駭客任務_中文字幕.srt
└── subtitles/                         (originals preserved)
    ├── Matrix_EN_Sub.srt
    └── 駭客任務_中文字幕.srt

Beyond matching, SubX converts between subtitle formats (SRT, ASS, VTT, SUB), corrects timing drift using local Voice Activity Detection, detects file character encodings, and translates subtitle cue text into another language using the configured AI provider while preserving timing and cue order.

Quick Start

# Install
curl -fsSL https://raw.githubusercontent.com/jim60105/subx-cli/master/scripts/install.sh | bash

# Configure (OpenRouter with free DeepSeek model)
export OPENROUTER_API_KEY="<YOUR_API_KEY>"
subx-cli config set ai.provider openrouter
subx-cli config set ai.model "deepseek/deepseek-r1-0528:free"

# Preview what SubX will do
subx-cli match --dry-run --copy /path/to/media/

# Execute
subx-cli match --copy /path/to/media/

For OpenAI, set OPENAI_API_KEY instead. For Azure OpenAI, set AZURE_OPENAI_API_KEY and AZURE_OPENAI_ENDPOINT. See the Configuration Guide for all provider options and environment variables.

Run with a local LLM (Ollama, LM Studio, llama.cpp, vLLM)

Prefer to keep subtitles on-device? Point SubX at any OpenAI-compatible local runtime — Ollama, LM Studio, llama.cpp's llama-server, vLLM, or any other server that speaks the POST /chat/completions protocol — by selecting the local provider. ollama is accepted as an alias and is normalized to local at config write time. No API key is required for typical local setups; SubX contacts only the configured base_url and ignores hosted-provider env vars (OPENAI_API_KEY, OPENROUTER_API_KEY, AZURE_OPENAI_*) when local is selected.

# Ollama running on the default port
subx-cli config set ai.provider local
subx-cli config set ai.base_url "http://localhost:11434/v1"
subx-cli config set ai.model "llama3.1:8b-instruct"

The endpoint can be loopback, a LAN host (http://192.168.x.x:port/v1), a tailnet (https://host.tailnet.ts.net/v1), or any reachable OpenAI-compatible URL. Both http:// and https:// schemes are valid for local. See the Local / Offline LLM Provider section for per-runtime examples and known compatibility limits.

Commands

The match, convert, sync, detect-encoding, and translate commands support -i for multiple inputs and --recursive for subdirectory scanning. See the Command Reference for full options, examples, and workflows.

Installation

Linux / macOS

# One-line installer (auto-detects host OS, architecture, and libc)
curl -fsSL https://raw.githubusercontent.com/jim60105/subx-cli/master/scripts/install.sh | bash

# Or download the binary directly (pick the asset matching your host)
curl -L "https://github.com/jim60105/subx-cli/releases/latest/download/subx-linux-x86_64" -o subx-cli
chmod +x subx-cli && sudo mv subx-cli /usr/local/bin/

Supported Release Targets

The installer auto-detects the host operating system and CPU architecture and downloads the matching pre-built binary from the latest GitHub Release.

From Source (any platform)

# From crates.io
cargo install subx-cli

# Or compile from the repository
git clone https://github.com/jim60105/subx-cli.git
cd subx-cli
cargo build --release

Supported Formats

Scripting & Automation

Pass --output json (before the subcommand) or set SUBX_OUTPUT=json to switch every covered subcommand to a stable, versioned JSON envelope on stdout — ideal for shell scripts, CI pipelines, and third-party tooling. Progress bars and status symbols are suppressed in JSON mode; the existing exit codes are preserved across both modes.

# Extract the first match candidate's confidence with jq
subx-cli --output json match --dry-run ./media \
  | jq -r '.data.candidates[0].confidence'

See Machine-Readable Output for the full envelope schema, error categories, per-command payloads, and scripting recipes. generate-completion is the only subcommand that explicitly rejects JSON mode.

Documentation

• Command Reference — full options, examples, and workflows for every subcommand
• Configuration Guide — all settings, environment variables, and troubleshooting
• Machine-Readable Output — the --output json contract for scripting and automation
• Technical Architecture — codebase structure and design decisions
• AI Provider Integration — how to add a new AI provider

License

GPLv3

GNU GENERAL PUBLIC LICENSE Version 3

Copyright (C) 2025 Jim Chen Jim@ChenJ.im.

This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.

You should have received a copy of the GNU General Public License along with this program. If not, see https://www.gnu.org/licenses/.