StepFun’s StepAudio 2.5 Realtime tops voice AI benchmarks in April 2026

Deepgram's secure voice AI deployment could revolutionize data-sensitive sectors by enabling advanced AI use without compromising privacy. The post Deepgram partners with Fortanix and Nvidia for secure voice AI deployment in regulated industries appeared first on Crypto Briefing.

The startup's own stack for Africa and Middle East is now handling more than 17,000 calls per day.

StepFun releases Step 3.7 Flash, a 198B MoE model with native vision, 256k context, and Advisor Mode. The post StepFun Releases Step 3.7 Flash: A 198B MoE Vision-Language Model for Coding Agents and Search Workflows appeared first on MarkTechPost.

The Shanghai lab that builds LLMs that punch above their weight just turned that same energy on voice—and the results are hard to ignore.

StepFun, the Shanghai-based AI lab, released StepAudio 2.5 Realtime in May 2026 — an end-to-end real-time speech large language model with fully customizable persona capabilities. The model connects via a WebSocket API, supports Chinese and English, and ranked first across all five benchmark dimensions tested in April 2026, including an 80.41 human evaluation score and 82.18 on paralinguistic comprehension. The post StepFun Releases StepAudio 2.5 Realtime: An End-to-End Voice Model with Roleplay-Specific RLHF and Paralinguistic Comprehension appeared first on MarkTechPost.

Wispr Flow says growth accelerated in India after its Hinglish rollout, even as voice AI products continue to face challenges.

OpenAI has introduced a Trusted Contact feature for ChatGPT that allows users to designate a friend or family member to receive automated alerts if conversations suggest self-harm risk. The company said human reviewers aim to assess safety notifications within one hour before deciding whether to contact the designated person via email, text, or in-app message. […]

Voice AI has a dirty secret. Most text-to-speech systems sound fine — until they don’t. They can read a sentence. What they cannot do is mean it. The rhythm is off. The emotion is flat. The speaker sounds like themselves for two seconds, then drifts into generic synthetic territory. That gap between intelligible audio and […] The post Closing the ‘Expressivity Gap’: How Mistral’s Voxtral TTS is Redefining Multilingual Voice Cloning with a Hybrid Autoregressive and Flow-Matching Architecture appeared first on MarkTechPost.