Tag: speech recognition
Granola adds dictation and meets Wispr in the middle
The meeting notetaker now types across any Mac app for free, months after Wispr's dictation tool added meeting…
Deepgram buys Daily to keep Pipecat outside its own walls
The speech company is absorbing the maintainer of the most widely used open framework for stitching voice agents…
Nvidia folds three speech jobs into a single open model
Nemotron 3.5 Transcribe handles 40 languages with 30B parameters and only about 4B active per step
Mistral bundles listening, thinking and speaking into one voice stack
Voxtral 3 pairs transcription and speech synthesis with a LiveKit-based agent SDK, priced per minute and runnable on…
Microsoft’s streaming speech model transcribes while you talk
Microsoft released MAI-Transcribe-2-Streaming, a real-time speech model it says tops Artificial Analysis accuracy while returning draft text in…
An open model sorts eight overlapping speakers in real time
NVIDIA released Nemotron 3 Diarization, a small open model that tops VoiceArena's leaderboard while running live conversation streams.
SpaceXAI doubles speech accuracy while holding its transcribe price
Grok Voice Transcribe 2.0 keeps batch pricing at $0.10 per hour of audio and claims the top accuracy…
OpenAI prices full-duplex voice at 5 cents a minute
GPT-Live-1 reaches the API, letting developers put natural turn-taking speech into products at a per-minute price.
Microsoft claims a 60-language speech record for MAI-Transcribe-2
The model posts a 5.2% average error rate on FLEURS and prices at $0.10 per hour of audio.
Meta folds transcription, speaker ID and silence detection into one model
Muse Voice Transcribe handles streaming speech-to-text, 20-plus speaker diarization, and endpointing in a single pass.
Gemini 3.5 Transcribe trims Google speech errors to 2.6 percent
Google's new speech model posts a 2.6 percent average error rate across 85-plus languages.