I would love to recommend it already, but my self-built privacy oriented real-time transcription service isn't ready for release yet. You can join my email wait

I would love to recommend it already, but my self-built privacy oriented real-time transcription service isn't ready for release yet. You can join my email waiting list [1], though. I tried to discover something that would stand the scrutiny of a German "Ausschreibung", which is a government call for bids. They require things to be GDPR-compliant by law, and they recently disqualified companies for using on-prem solutions by US Cloud providers with the argument that if it's a US company, the NSA can still force them to disclose GDPR-protected data. I reviewed all offerings I could find on the market. I couldn't find anything good, which is when I decided to build my own. For what it's worth, I also published a paper on improving German ASR already [4]. In my opinion, the only 100% offline and, hence, 100% private solutions that are ready to use right now are VOSK and Coqui STT. But their recognition quality is atrocious. Like pre-2016 word-error-rate on LibriSpeech / CommonVoice. Then there's Scribosermo, which is converted NVIDIA models, so it comes with a lot of rules for how you're allowed to use it, and quality is still worse than all US clouds and it's quite slow, like 2s delay. In addition to that, I could only find companies who more or less openly re-sell the cloud AI solutions from Microsoft, Amazon, Google, or Alibaba. But that won't fly for German government bids [2]. I believe your best option is to wait for me ;) to finish my local tool [1] which will have a WebRTC server so that you can just script it however you want :) And your second best option is probably to use HuggingFace Transformers with Facebook's pre-trained base models [3]. [1] https://madmimi.com/signups/f0da3b13840d40ce9e061cafea6280d5/join https://madmimi.com/signups/f0da3b13840d40ce9e061cafea6280d5... [2] https://gdprhub.eu/index.php?title=VK_Baden-W%C3%BCrttemberg_-_1_VK_23/22 https://gdprhub.eu/index.php?title=VK_Baden-W%C3%BCrttemberg... [3] https://huggingface.co/blog/fine-tune-wav2vec2-english https://huggingface.co/blog/fine-tune-wav2vec2-english [4] https://paperswithcode.com/paper/tevr-improving-speech-recognition-by-token https://paperswithcode.com/paper/tevr-improving-speech-recog...