StepFun Releases Step-Audio 2 mini, a Unified Audio AI
The new open-source model handles both speech recognition and audio generation in a single, end-to-end architecture.
AI research company StepFun has released Step-Audio 2 mini, a new audio language model designed for both understanding and generating human speech. The model is presented as a compact, end-to-end system, suggesting an efficient and unified architecture.
Unlike traditional pipelines that use separate components for automatic speech recognition (ASR) and text-to-speech (TTS), Step-Audio 2 mini integrates these capabilities. This approach aims to simplify the development of voice-enabled applications by handling the entire audio processing loop within a single framework.
The 'mini' in its name suggests it is a smaller variant, likely optimized for resource efficiency, though a specific parameter count was not disclosed. By releasing the model under the permissive Apache 2.0 license, StepFun is making it available for a wide range of academic and commercial projects.
Step-Audio 2 mini joins a growing field of open, multimodal models that seek to process and generate data across different formats. Interested developers can explore the model and its capabilities on the Hugging Face Hub.
Sources
- Visit
stepfun-ai/Step-Audio-2-mini
Hugging Face
More from StepFun
All StepFun releases →
StepFun Posts Step-5-Preview Vision-Language Model
The Chinese AI lab's early look at its next multimodal system arrives on Hugging Face ahead of a full release.
StepFun's StepAudio 3 Realtime targets live voice AI
The audio-language foundation model builds a listen-converse-think-act loop aimed at natural, low-latency spoken interaction.
StepFun's StepAudio 3 Music Plans Before It Plays
The new open model separates musical structure from sound, generating long-form tracks from text prompts with an explicit planning stage.
More in Any-to-Any
All Any-to-Any →
Liquid AI's LFM3-Omni-2B brings any-to-any AI to the edge
The 2-billion-parameter model combines vision, audio and speech-to-speech in a package small enough to run on device.
MiniMax Debuts M3.1 Flash, a Multimodal MoE Model
The new release pairs a mixture-of-experts design with agentic, coding, and video-understanding abilities in a single open model.
IBM's Granite Switch 4.1 Unifies Chat, Logic, Speech
A single endpoint now routes text conversation, decision-making, and speech synthesis across IBM's Apache-licensed Granite line.
0 comments
No comments yet. Be the first to weigh in.