StepAudio3 Series Model Launch: Five Models Simultaneously Released on the Open Platform

StepZenith officially launched the StepAudio3 series of models today, including StepAudio3Realtime, StepAudio3ASR, StepAudio3TTS, StepAudio3Gen, and StepAudio3Music. All five models have been released on the StepZenith Open Platform, with multiple models ranking first globally in the Artificial Analysis rankings. This series covers scenarios such as human-level voice generation, complete audio generation, real-time voice interaction, complex voice understanding, and music creation.

Among them, StepAudio3Realtime supports native full-duplex real-time conversation, capable of understanding semantics, tone, emotion, paralanguage, and ambient sounds simultaneously, and supports parallel reasoning and voice generation, Tool Call, and asynchronous execution of long tasks. It achieved a comprehensive score of 98.9% in the Artificial Analysis Conversational Dynamics ranking, ranking first globally; its voice reasoning accuracy reached 99.7%, also ranking first globally.

StepAudio3ASR integrates speech recognition and large model context understanding capabilities, supporting Chinese, English, dialects, mixed Chinese and English, long audio, and professional scenarios such as medical, legal, and financial fields, with a WER of only 1.7%, tied for first globally. StepAudio3TTS enhances vocal tone, intonation, rhythm, pauses, and paralanguage expressions such as laughter, hesitation, and stuttering, and supports streaming generation, targeting real-time voice interaction.

Additionally, StepAudio3Gen can uniformly generate character voices, sound effects, ambient sounds, and background music, and supports multi-character and voice timing control; StepAudio3Music supports song creation, a cappella accompaniment, cover songs, and multi-round creation based on ABC notation, allowing natural language control over musical style, melody, rhythm, instruments, and emotion. Overall, StepAudio3 is expanding the capabilities of voice models from single recognition or generation to complete audio content production and real-time intelligent interaction.

StepZen Launches StepAudio 3 Series Voice Large Model, Strongly Secures Multiple Global Firsts!

StepFun launched StepAudio3 voice models on Sept 15, covering Realtime, ASR, TTS, Gen, and Music. Several ranked first on Artificial Analysis. All five are live on the open platform for lifelike voice, full audio generation, real-time interaction, complex speech understanding, and music creation.....

Canadian researcher calls for indefinite pause in cutting-edge AI development: human extinction is a matter of time otherwise

Canadian AI researcher Krueger warns that slowing AI development is far from enough to avert human doom; frontier AI development must be paused immediately and indefinitely. He is an assistant professor at University of Montreal and a core member of Mila, focusing on reducing AI extinction risk.....

Data Security Sounds the Alarm! Major Companies Ban Cutting-Edge AI, Why Are Giants Like NVIDIA Rushing to Return to Their Own Models?

Several major enterprise customers (NVIDIA, Palantir, Booz Allen Hamilton, etc.) have restricted the use of Anthropic and OpenAI's cutting-edge models in sensitive projects due to concerns over proprietary data security, triggering a shockwave in the enterprise AI market. The cause was Anthropic's policy announced in June this year, which allowed its Fable model to retain customer data for 30 days to detect abuse, triggering a trust crisis.

WeChat Responds to Privacy Controversy Surrounding Weixiao AI: It Will Not Read Users' Deleted Moments

The WeChat AI assistant 'Weixiao' has sparked controversy over privacy concerns and has topped the hot search list. WeChat clarified that claims about directly reading chat records or spying on privacy are inaccurate. The assistant is a native AI, currently in a limited beta test, supporting text and voice interaction, capable of checking weather, express delivery, and takeout. It can also proactively trigger in scenarios such as Moments, official accounts, chats, and short video sections to assist with content summarization.

Leaked: Anthropic, OpenAI, and Google Secretly Discussing the Establishment of an AI Safety Standards Organization

Sources say Anthropic, OpenAI, and Google had begun talks to jointly establish an AI industry standards body before Anthropic's CEO publicly called for it. Since July, working groups below CEO level from the three companies have met regularly. OpenAI CEO Altman also internally supported creating an AI testing and auditing body.....