AudioToText

AudioToText is an open source project that transcribes and translates audio files using Whisper from OpenAI together with the DeepL translator. It runs as a notebook in Google Colab, where a cloud GPU is assigned to execute the code, so no local machine with an accelerator is required. A local installation and a command line version are also available for people who prefer to run everything on their own hardware. The project suits podcasters, video editors, researchers and anyone producing captions. Whisper handles transcription in English and in a long list of other source languages, with automatic detection or manual selection of the language, and it can translate any supported language into English. Because Whisper does not translate into languages other than English, the project routes transcriptions through the DeepL API when another target language is needed. Results can be saved as TXT, VTT, SRT, TSV and JSON, so subtitle files come straight out of the same run. Users choose between the open source Whisper models and the hosted API, and the repository ships worked examples alongside the notebook. No pricing is published. The project is distributed on GitHub with downloadable archives and carries no plans, subscription or paid tier. Running it in Google Colab or locally is free, although choosing the API option rather than the open source models draws on accounts you hold directly with the respective providers.