9 October 2024<\/p>\n
<\/p>\n
Donate your voice<\/strong> via the platform\u00a0Mozilla Common Voice<\/a>\u00a0(Instructions<\/a>)<\/strong> and help us reach 600 hours of recorded speech!<\/p>\n If we use your voice in the training, you will be more easily recognisable in the new version of Buki. We most need voices of children and older people, because they are the least represented in the training data.<\/p>\n <\/p>\n To ensure the speed and accuracy of Buki, the KANTAROT Foundation will provide free testing over the next two weeks, using the Hugging Face platform. The models will be open for use.<\/strong> From 9 October at 14:30 Macedonian time, at the following link<\/a><\/strong>:<\/p>\n https:\/\/huggingface.co\/Macedonian-ASR<\/strong><\/a><\/p>\n <\/p>\n You will have access to the Buki-wav2vec2 model (fast and accurate), Buki-Whisper (slower and more accurate), as well as a comparison between Buki-Whisper and the open models from OpenAI and Meta. Next link<\/a><\/strong> has Video demonstration of how to test the model<\/strong>.<\/p>\n We hope to get hold of a powerful graphics card soon so that we can continue using Buki locally. In the meantime, please donate a vote, because who donates<\/a>\u00a0Transcribe!<\/p>\n <\/p>","protected":false},"excerpt":{"rendered":" 9 October 2024 Donate your voice via the Mozilla Common Voice platform (instructions) and help us reach 600 hours of recorded speech! If we use your voice in the training, you'll be more easily recognisable in\u2026