Topic
Speech generation, voice agents, speech-to-text, and real-time audio models.
Latest Signals
A Hacker News thread discusses free text-to-speech options, highlighting two points but no comments. The conversation centers on accessible TTS tools available at the time.
A natural field experiment with 70,000 job applicants compared AI voice agent interviews to human recruiter interviews, with humans making final hiring decisions in both cases. The study examines whether AI can reduce variance in information collection and improve organizational outcomes.
Inflect TTS v2+ONNX offers 9 million and 4 million parameter text-to-speech models that run directly in the browser. This enables efficient, client-side voice synthesis without server dependency.
A new duplex speech model designed for revenue calls has been introduced, enabling real-time, interactive voice communication. The model was discussed on Hacker News, highlighting its potential applications in business contexts.
FlashRT is an agent harness designed to optimize deployment of real-time multimodal applications like voice agents and interactive video generation by managing model pipelines with application-specific placement and parallelism. It addresses limitations of existing serving systems that rely on fixe...
New research shows that AI voice synthesis and large language models can automate voice phishing attacks, removing the need for human operators. A large-scale study assessed U.S.
The study evaluates the reliability of Gemini models (2.5 Flash, 3.5 Flash, 3.1 Pro) as audio judges scoring full-duplex voice agent conversations from raw stereo waveforms. Gemini 2.5 Flash was validated against human raters across 209 sessions on eight production dimensions.
DETECT-3B-Omni is a GDPR-compliant deepfake audio detector that bases its decisions on acoustic artifacts rather than speech content or speaker identity. A large-scale study using 10,240 samples from diverse US English speakers across 30 states and 8 AI voice-cloning systems confirms its semantic i...
Hugging Face and Cerebras have introduced Gemma 4, a model designed to enhance real-time voice AI applications. This collaboration aims to improve the performance and responsiveness of voice-based AI systems.
Advanced AI voice cloning technologies raise significant legal and ethical issues about protecting vocal identity. The similarity between OpenAI's ChatGPT-4o voice and Scarlett Johansson's highlights these concerns.
Speech generation, voice agents, speech-to-text, and real-time audio models.