4 filings, one each in 4 layers.
Code-switch without language flags at $0.21/hr, plus a prompt that steers entity recognition. Named customers: Granola, Fireflies, Dovetail.
AssemblyAI brought Universal-3.5 Pro to pre-recorded audio with native code-switching across 18 languages and contextual prompting, priced at $0.21/hr.
Universal-3.5 Pro
GA for the conversational TTS SKU. Pricing and plan mapping were not in the source we have (looks like LinkedIn rather than a dedicated blog).
ElevenLabs moved Eleven v3 Conversational, its realtime expressive TTS model, to general availability for production streaming voice apps across 70+ languages.
Eleven v3 Conversational
Eleven v3 Conversational, our most expressive model for realtime speech, is now generally available. For developers building voice experiences that respond with real emotion, Eleven v3 Conversational includes audio tags for fine-grained control and support across 70+ languages.
https://x.com/ElevenLabs/status/2090136227617952145
Avatar session param passthrough.
livekit-agents 1.7.0 forwards avatar session extra params through Spatius (1.0.5).
livekit-agents 1.7.0
Observability privacy controls distinct from the expressive-mode blog row, covering Python and Node. Class-2: folded quiet patch(es) 1.7.1 into this row.
livekit-agents and @livekit/agents 1.7.0 semantically redact detected entities from chat history and audio recordings and filter sensitive data from logs and traces; the Node core package lands on the same 1.7 line.
livekit-agents / @livekit/agents 1.7.0