Advanced Options#

Scripting#

Since version 0.7, noScribe has a command-line interface and can be integrated into your own scripts. noScribe.exe --help shows the available options. With --no-gui, the main window is suppressed and noScribe runs fully in command-line mode.

Configuration#

After the first start, the file config.yml is located in the user profile:

• Windows: C:\Users\<username>\AppData\Local\noScribe\noScribe\config.yml
• macOS: ~/Library/Application Support/noscribe/config.yml
• Linux: ~/.config/noScribe/config.yml

There you can adjust the UI language and other options. Note: Only edit this file when noScribe is not running, otherwise your changes will be overwritten.

Logs#

In the same profile directory there is a log folder with detailed information about each transcript - useful for troubleshooting.

Use Special Transcription Models#

noScribe comes with two transcription models (precise/fast), both based on Whisper V3 Turbo. You can also use other Whisper models that are better tuned for specific languages.

Installation:

1. In noScribe under Model, open the list and select Add AI models. This opens the whisper_models folder in the noScribe config directory (paths above).
2. Create a separate subfolder for each model. The name of this folder will be the model name shown in noScribe.
3. Place all files belonging to the model in that folder (typically: config.json, model.bin, preprocessor_config.json, tokenizer.json, vocabulary.json).
4. The new model should now appear as an additional selection under Models in noScribe.

Notes:

• Only models in Faster-Whisper or CTranslate2 format are supported. Models in the original OpenAI Whisper format can be converted.
• Some downloads include ctranslate2_ prefixes - remove these (e.g., ctranslate2_vocabulary.json -> vocabulary.json).

Models:

• The official Faster-Whisper repository (15 models): https://huggingface.co/Systran This also includes the Whisper-v2 model, which was the default up to version 0.5 of noScribe and is still preferred by some, even though it is significantly slower. The English-only models (ending in “.en”) are also interesting because they run noticeably faster.
• A French-tuned version of Whisper V3: https://huggingface.co/bofenghuang/whisper-large-v3-french/tree/main/ctranslate2 (If the downloaded files have the prefix “ctranslate2_”, please remove it.)
• A German-optimized version of Whisper V3 (my experience: only slight improvements in transcription quality, but less precision in pause detection and in identifying overlapping speech sequences): https://huggingface.co/aseifert/faster-whisper-large-v3-turbo-german/tree/main You also need the file tokenizer.json from the original Whisper V3 repository: https://huggingface.co/Systran/faster-whisper-large-v3/tree/main

You can find more models on huggingface. Search for “faster-whisper” and the desired language (or just “whisper” to find models that can be converted).

I use this version of Whisper V3 Turbo as the “precise” model: https://huggingface.co/mobiuslabsgmbh/faster-whisper-large-v3-turbo

If you find a good new model, please share it in this discussion: https://github.com/kaixxx/noScribe/discussions/132.

Contributing to Development#

• noScribe is based on Python 3.13.
• The AI models (Whisper) are too large for GitHub. They must be downloaded separately if you want to run noScribe from source. See the notes in the models folder.
• Tests, bug reports, and pull requests are welcome. I try to respond quickly. If not, feel free to remind me politely. Please understand that I cannot accept every idea.

Translations#

The UI is translated into many languages. However, some of these translations were created by machine and may contain errors. I appreciate corrections.

The language files are in the trans folder. Changes should follow YAML conventions. Switching the UI language is done via the locale setting in config.yml.