Common Voice
Mozilla crowdsourced multilingual speech dataset
Episodes
-
#6192: What Actually Happens When You Train a ModelData in, model out — except the "data" is pairs, the split is the whole game, and one leaked speaker can fake your score. -
#6032: Blending Datasets vs Merging Weights for Fine-TuningMerging weights happens in parameter space. Blending datasets happens in data space — and the messy part is worth it. -
#6022: Fine-Tuning for Code-Switched Hebrew on 2 Hours of AudioTwo hours of one man's voice beat models trained on 120,000 hours — but only for code-switched Hebrew. -
#5962: How Word Error Rate Lies to UsWER is the number everyone reports for speech recognition — and the one that agrees with human judgment least. Here's what it actually measures. -
#5485: Fine-Tuning Parakeet for Hebrew and Your Own JargonNVIDIA's Parakeet beats Whisper on Android — but can you teach it Hebrew, or just your own jargon? Two answers, one much happier. -
#4641: Mozilla's Hidden Projects: Beyond the BrowserMozilla is more than just Firefox. Discover Common Voice, Monitor, Thunderbird, and Rust — the public infrastructure projects shaping the open web. -
#1810: Why Your TTS Sounds Great in English, Terrible Everywhere ElseEnglish AI voices are polished, but global languages hit a wall. Here's why text-to-speech breaks down for Hebrew, Hindi, and beyond. -
#272: The Bill is Due: AI Training and Intellectual PropertyCan you "untrain" an AI? Herman and Corn explore the legal and technical battle over copyrighted data and the future of machine unlearning. -
#9: Benchmarking Custom ASR Tools - Beyond The WERBenchmarking custom ASR fine-tunes: We're diving deep beyond the WER to truly measure performance.