{"slug": "kyutai-stt", "name": "Kyutai STT", "category": "ai-models", "category_name": "AI models", "description": "", "recommenders": 3, "commenters": 3, "warners": 0, "score": 2.547, "rising": -0.98, "first": "2026-02-05", "latest": "2026-02-27", "why": "I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt", "why_is_quote": true, "has_verdict": false, "sentiment": {"recommend": 3, "neutral": 0, "compare": 0, "warn": 0}, "spark": [0, 0, 0, 3, 0, 0, 0, 0, 0, 0, 0, 0], "listed": true, "loved": [], "caveat": "", "short_quote": "I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt", "kind": "", "icon": true, "domain": "kyutai.org", "github_repo": null, "official_url": "https://kyutai.org/stt", "pricing": "", "aliases": ["Kyutai STT"], "verdict": null, "praise": [{"label": "", "summary": "", "quotes": [{"id": 754890400, "item_id": 47180650, "url": "https://news.ycombinator.com/item?id=47180650", "by": "ahaferburg", "date": "2026-02-27", "version": null, "stance": "recommend", "quote": "which is very low latency", "excerpt": "Agree about the latency requirement. There's https://kyutai.org/stt, which is very low latency. But it seems not as hackable.", "context": "", "use_case": "speech-to-text", "switched_from": "", "story": "Parakeet.cpp \u2013 Parakeet ASR inference in pure C++ with Metal GPU acceleration"}, {"id": 751357504, "item_id": 46959844, "url": "https://news.ycombinator.com/item?id=46959844", "by": "sofixa", "date": "2026-02-10", "version": null, "stance": "recommend", "quote": "it should work on edge devices, Macs and iPhones", "excerpt": "https://kyutai.org/stt has an implementation for MLX and mentions iPhones, so it should work on edge devices, Macs and iPhones.", "context": "", "use_case": "", "switched_from": "", "story": "Rust implementation of Mistral's Voxtral Mini 4B Realtime runs in your browser"}, {"id": 750438049, "item_id": 46902378, "url": "https://news.ycombinator.com/item?id=46902378", "by": "espadrine", "date": "2026-02-05", "version": null, "stance": "recommend", "quote": "I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt", "excerpt": "It is quite impressive. I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt If I look at the architecture of Voxtral 2, it seems to take a page from Kyutai\u2019s delayed stream modeling. The reason the delay is configurable is that you can delay the stream by a variable number of audio tokens. Each audio token is 80 ms of audio, converted to a spectrogram, fed to a convnet, passed ", "context": "", "use_case": "", "switched_from": "", "story": "Voxtral Transcribe 2"}]}], "caveats": [], "who": [], "tags": [], "use_cases": ["speech-to-text"], "compared": [], "makers": [], "threads": [{"url": "https://news.ycombinator.com/item?id=47176239", "title": "Parakeet.cpp \u2013 Parakeet ASR inference in pure C++ with Metal GPU acceleration", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46954136", "title": "Rust implementation of Mistral's Voxtral Mini 4B Realtime runs in your browser", "mentions": 1}, {"url": "https://news.ycombinator.com/item?id=46886735", "title": "Voxtral Transcribe 2", "mentions": 1}], "sentiment_by_month": {"2025-11": {}, "2025-12": {}, "2026-01": {}, "2026-02": {"recommend": 3}, "2026-03": {}, "2026-04": {}, "2026-05": {}, "2026-06": {}, "2026-07": {}, "2026-08": {}, "2026-09": {}, "2026-10": {}}, "versions": [], "version_unnamed": 3, "family_ref": null, "runs_on": null, "lines": [], "quotes": [{"id": 754890400, "item_id": 47180650, "url": "https://news.ycombinator.com/item?id=47180650", "by": "ahaferburg", "date": "2026-02-27", "version": null, "stance": "recommend", "quote": "which is very low latency", "excerpt": "Agree about the latency requirement. There's https://kyutai.org/stt, which is very low latency. But it seems not as hackable.", "context": "", "use_case": "speech-to-text", "switched_from": "", "story": "Parakeet.cpp \u2013 Parakeet ASR inference in pure C++ with Metal GPU acceleration"}, {"id": 751357504, "item_id": 46959844, "url": "https://news.ycombinator.com/item?id=46959844", "by": "sofixa", "date": "2026-02-10", "version": null, "stance": "recommend", "quote": "it should work on edge devices, Macs and iPhones", "excerpt": "https://kyutai.org/stt has an implementation for MLX and mentions iPhones, so it should work on edge devices, Macs and iPhones.", "context": "", "use_case": "", "switched_from": "", "story": "Rust implementation of Mistral's Voxtral Mini 4B Realtime runs in your browser"}, {"id": 750438049, "item_id": 46902378, "url": "https://news.ycombinator.com/item?id=46902378", "by": "espadrine", "date": "2026-02-05", "version": null, "stance": "recommend", "quote": "I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt", "excerpt": "It is quite impressive. I have seen the same impressive performance about 7 months ago here: https://kyutai.org/stt If I look at the architecture of Voxtral 2, it seems to take a page from Kyutai\u2019s delayed stream modeling. The reason the delay is configurable is that you can delay the stream by a variable number of audio tokens. Each audio token is 80 ms of audio, converted to a spectrogram, fed to a convnet, passed ", "context": "", "use_case": "", "switched_from": "", "story": "Voxtral Transcribe 2"}]}