Qwen Releases 0.6B Model for Audio-Text Alignment
The new open-source tool, based on the Qwen3 architecture, precisely synchronizes audio recordings with their corresponding text transcripts.
Alibaba's Qwen team has released a new specialized tool for audio processing, the Qwen3 ForcedAligner 0.6B. This compact 600-million-parameter model is designed for a specific and crucial task in speech AI: aligning existing text with an audio recording.
Unlike standard speech-to-text models that generate text from scratch, a forced aligner takes both an audio file and its transcript as input. It then determines the precise start and end times for each word in the audio, effectively synchronizing the two. This capability is essential for creating accurately timed subtitles, preparing high-quality datasets for training other speech models, and conducting phonetic research.
The model is built on the Qwen3 architecture and is available on the Hugging Face Hub under a permissive Apache 2.0 license, allowing for broad commercial use. Its relatively small size suggests it can be run efficiently, making this alignment technology more accessible to developers and researchers.
The release of Qwen3 ForcedAligner adds another foundational component to the open-source audio ecosystem, providing a key tool for building more sophisticated applications that handle spoken language.
Sources
- Visit
Qwen/Qwen3-ForcedAligner-0.6B
Hugging Face
More from Qwen · Alibaba
All Qwen · Alibaba releases →Qwen-Audio-3.1-ASR Adds Speaker-Aware Transcription
Alibaba's new open audio-language model transcribes and diarizes speech across 30 languages under an Apache 2.0 license.
Qwen-Image-2.1-Turbo targets faster generation
Alibaba's Qwen team ships a distilled, few-step variant of its image model built for speed.
Qwen-Image-2.1 Adds RGBA to Image Generation
Alibaba's Qwen team updates its open image model with transparency support and joint text-to-image and editing capabilities.
More in Speech → Text
All Speech → Text →IBM's Granite Speech 5.0 Adds Speaker Attribution
The Apache-2.0 speech model handles transcription, translation, and figuring out who said what.
NVIDIA's Nemotron 3.5 Transcribe Brings Promptable ASR
A sparse 30B mixture-of-experts model handles multilingual transcription, translation, and speaker diarization from a single system.
TII extends the Falcon family into speech recognition
The Abu Dhabi lab behind the Falcon language models releases Falcon ASR, an open speech-to-text system.
0 comments
No comments yet. Be the first to weigh in.