EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·24d ago·5 min read
Grok Voice Transcribe 2.0
TL;DR
Grok Voice Transcribe 2.0 is a highly accurate speech-to-text model, achieving twice the accuracy of its predecessor while maintaining the same cost. It excels in challenging audio environments, making it one of the top transcription models available.
✦ Why It Matters
Consider integrating Grok Voice Transcribe 2.0 into your applications for improved transcription accuracy in challenging audio conditions.
Key Takeaways
- Upgrade to Grok Voice Transcribe 2.0 for improved accuracy.
- Utilize automatic language detection for diverse applications.
- Integrate with existing APIs without code changes.
- Leverage accurate transcripts for new AI workflows.
Full Summary
Grok Voice Transcribe 2.0 is the latest iteration of a speech-to-text model designed to provide high accuracy in challenging audio environments. Built on the Grok Voice audio foundation model, it has been trained on a unique dataset that includes noisy, multilingual audio from various real-world settings.
This model significantly improves transcription accuracy, achieving a word error rate of just 6.8% for short phrases, down from 20.6% in the previous version. It ranks first among 32 streaming models on the public Artificial Analysis leaderboard and outperforms its predecessor across multiple internal evaluation sets, including customer support calls and multilingual voice commands.
The pricing remains the same as Grok Voice Transcribe 1.0, making it accessible for users. With automatic language detection and support for 19 languages, it is well-suited for global applications.