Open-source append-only streaming speech recognition model with configurable low-latency decoding, English and Mandarin.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
R2T2 (also referred to as R2D2 in some coverage) is an open-source streaming automatic speech recognition (ASR) model. According to its website, it is append-only, meaning committed transcript text is not revised, and it offers configurable low-latency decoding. The project publishes code, model weights, a technical report, and a hosted demo.
The site shows a side-by-side comparison with GPT-Live-Transcribe on the same audio, and replays of English and Mandarin clips with synchronized audio and committed transcripts. A YouTube creator (AI Search) described it as a roughly 4 GB model with lower word error rate and latency than other transcription models, including GPT Live Transcribe, in English and Chinese. This is a secondary claim and has not been verified here.
It is aimed at developers who want to run real-time transcription themselves or try the hosted demo. The vendor, license, and exact release date were not stated in the provided material.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Open-source real-time transcription model of about 4 GB. It shows the lowest word error rate and latency compared with other transcription models, including GPT Live Transcribe, in English and Chinese.
Closest neighbours by meaning, not just by category.
Model · Microsoft
Microsoft AI's streaming speech-to-text model for near real-time transcription, succeeded by MAI Transcribe 2.
Model
Speech recognition model trained for noisy, reverberant and distorted real-world audio, with RL post-training.
Model · Fermion Research
Open 164 MB English speech recognition model from Fermion Research, derived from Parakeet TDT 0.6B v3.
Model · Nyra Labs
Controllable multilingual speech recognition with verbatim or clean output and word-level timestamps, from Nyra Labs.
Model · Microsoft
Microsoft AI's second-generation speech-to-text model for transcribing noisy, domain-specific audio.
Model · IndexTeam
Open-weights zero-shot text-to-speech model from Bilibili's IndexTeam with voice cloning and emotion control in 5 languages.
