https://github.com/xkeyc/fl_caption
Offline real-time captioning software written in Flutter and Rust, powered by Whisper and LLM.
https://github.com/xkeyc/fl_caption
candle flutter flutter-desktop llm rust whisper
Last synced: 6 months ago
JSON representation
Offline real-time captioning software written in Flutter and Rust, powered by Whisper and LLM.
- Host: GitHub
- URL: https://github.com/xkeyc/fl_caption
- Owner: xkeyC
- License: gpl-3.0
- Created: 2025-03-12T15:56:51.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-04-01T01:11:36.000Z (6 months ago)
- Last Synced: 2025-04-01T02:22:22.320Z (6 months ago)
- Topics: candle, flutter, flutter-desktop, llm, rust, whisper
- Language: Dart
- Homepage:
- Size: 10.6 MB
- Stars: 49
- Watchers: 1
- Forks: 2
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# FL Caption
[](https://github.com/xkeyC/fl_caption/actions/workflows/windows_build.yml)离线实时字幕软件,使用 Flutter 和 Rust 编写,由 LLM 和 基于 candle 推理框架的 Whisper 驱动。
演示视频:https://www.bilibili.com/video/BV1VyQtYMEWA
QQ群:1037016702

## 使用说明
1. 通过 Release 下载压缩包并解压:https://github.com/xkeyC/fl_caption/releases
2. 首次使用请点击设置图标,选择合适的语音模型,并点击下载按钮
3. 下载成功后请选择语音语言与字幕语言,并设置 llm api 信息,完成后点击保存
4. 字幕应当会正常开始运行。
## 常见问题
1. 模型下载无进度或下载失败:请尝试参照 https://github.com/xkeyC/fl_caption/issues/1 设置 `HF_ENDPOINT` 环境变量,或打开 https://github.com/xkeyC/fl_caption/blob/main/lib/common/whisper/models.dart 的链接手动下载文件,文件名即为 name 的值,如 `base` `large-v3_q4k` 等。
2. 启动后卡在 Wait for Whisper:如果开启了 CUDA 加速,请检安装 CUDA Toolkit (下载链接:https://developer.nvidia.com/cuda-downloads?target_os=Windows ) 。 若不是 CUDA 的问题,请开一个 issue 并说明硬件规格。
> 提示:安装时仅选择 Development 和 Runtime -> Libraries 可优化安装速度与文件体积。
