| --- |
| title: CraftMann |
| emoji: 🧰 |
| colorFrom: gray |
| colorTo: green |
| sdk: static |
| pinned: false |
| --- |
| |
| # CraftMann |
|
|
| **Well-made tools for people who make things.** CraftMann builds focused, free, open-source desktop apps. Our first app is [CaptureMann](https://github.com/craft-mann/capturemann): screenshots and screen and audio recording, with local AI for transcripts, captions, speaker labels and meeting minutes that stay on your computer. |
|
|
| ## What lives here |
|
|
| Model files that CaptureMann can download for its local AI. Nothing is downloaded unless you press **Download** in the app, and nothing you record is ever sent here. |
|
|
| | Repo | What it is | |
| |---|---| |
| | `phowhisper-ggml` | [PhoWhisper](https://github.com/VinAIResearch/PhoWhisper) (VinAI Research), Vietnamese fine-tunes of Whisper, converted to the ggml format that whisper.cpp reads. Sizes: small, medium, large. | |
|
|
| ## About the converted models |
|
|
| - The weights are VinAI Research's work (`vinai/PhoWhisper-*`, BSD-3-Clause). We only convert the file format; the model is unchanged. Please cite and credit the original authors. |
| - Each repo's README lists the exact conversion steps, the tool versions, and the SHA-256 of every file. The app checks that SHA-256 before it uses a file. |
| - Speech recognition quality depends on your audio and language. We publish measurements, with the data and settings used, in the CaptureMann repository under `verification/benchmarks/`. |
|
|
| ## Links |
|
|
| - App and source: [github.com/craft-mann/capturemann](https://github.com/craft-mann/capturemann) (Apache-2.0) |
| - Website: [craftmann.app](https://www.craftmann.app) |
| - Contact: [LinkedIn](https://www.linkedin.com/company/craftmann) |
|
|
