Voice AI Weekly #16 - 2nd Week of November 2026
This week saw a series of contact center acquisitions and platform reorganizations. There were also notable technology announcements, such as an API that makes direct determinations from voice signals without converting them to text, and STT tailored for agent call environments. Here is a summary of the five key stories highlighted by MAGO.
1. Hume Releases Audio Decision API to Query Audio Directly
Hume released the Audio Decision API on November 2. When a multiple-choice question is submitted alongside an audio file, the model directly returns the probability values for each option without requiring text transcription or separate fine-tuning. Hume stated that its agreement rate with the human evaluator panel average is 70.7%, outperforming individual annotators (56.5%), with a median latency of around 0.23 seconds. Currently, it is only available via waitlist registration, and detailed pricing has not been disclosed. Technology that directly assesses state from voice signals themselves—bypassing the text conversion process—is increasingly taking shape as global commercial products.
2. Soniox STT v6 Features Agent Speech Filtering
Soniox announced its new real-time (stt-rt-v6) and asynchronous (stt-async-v6) STT models on November 3. The system is designed to receive the text spoken by the voice agent over the same WebSocket connection as the user audio, applying echo cancellation technology to prevent the agent's playback audio from bleeding into transcription results. Benchmark figures and pricing were not disclosed, and support for the legacy v5 model will end on December 15. The focus of STT technology is shifting beyond simple dictation accuracy toward understanding conversational context and call environments.
3. Five9 Acquires Level AI for Approximately $325 Million
Five9 announced on November 5 that it will acquire contact center AI company Level AI for approximately $325 million in cash. Level AI possesses conversational intelligence technology that processes over 1 billion customer interactions annually, and the two companies have maintained a partnership since 2022. On the same day, Five9 also reported its Q3 revenue of $323.3 million (13% year-over-year growth). Following Salesforce's acquisition of Cresta last week, major CCaaS platforms are clearly making moves to internalize conversational analytics capabilities.
Source: Five9 SEC Filing Press Release
4. ElevenLabs Natively Integrates with Four Major CCaaS Platforms
ElevenLabs announced on November 4 that its voice agent solution, ElevenAgents, has been natively integrated with NICE CXone, Genesys Cloud, Amazon Connect, and Five9. Adopting enterprises can deploy ElevenLabs agents while preserving their existing call routing systems, and conversation context is maintained seamlessly when transferring calls to live agents. In enterprise voice AI adoption, seamless interoperability with legacy systems appears to be just as critical as the model's raw specifications.
Source: ElevenLabs Official Blog
5. SoundHound AI Cuts 27% of Workforce Amid 48.5% Q3 Revenue Growth
SoundHound AI announced on November 5 that it recorded Q3 revenue of $62.4 million, up 48.5% year-over-year. However, through a regulatory filing on the same day, the company revealed it is laying off approximately 27% of its total workforce, anticipating pre-tax charges of $18 million to $22 million, including severance payments. The company provided a full-year 2026 revenue guidance of $250 million to $260 million and maintained its target of reaching profitability in the second half of 2027.
Source: SoundHound AI SEC Form 8-K
If you have any questions or are considering speech emotion recognition, domain-specific STT, or on-premise voice AI deployments, please feel free to reach out to contact@holamago.com anytime. We will be back next week with more carefully curated news worth reviewing from the field.