2022 was another busy year at Picovoice with new products, languages, SDKs, team members and customers. See the 2022 highlights from Picovoice and stay tuned for the 2023 announcements, starting in January.
Product Updates
Even Better Wake Word and Intent Detection
2.1 Releases of Porcupine Wake Word and Rhino Speech-to-Intent engines achieve even higher accuracies than alternative wake word and natural language understanding engines. Plus, Rhino has new capabilities, such as manual end-point adjustments to improve the end-user experience and macros to improve the developer experience.
Introduction of Local Speech-to-Text with Cloud Accuracy
Leopard Speech-to-Text and Cheetah Speech-to-Text for streaming are new additions to Picovoice’s portfolio. These small and efficient speech-to-text models achieve cloud-level accuracy on the device. They’re better, faster and stronger.
Improving Developer Experience
Picovoice Console has got new user interface to streamline the developer experience. More changes to improve the experience of designing and training speech models are on the way.
SDK Updates
Web SDKs for All Engines
All Picovoice engines run within web browsers without sending data to a 3rd party cloud with the release of Web SDKs for speech-to-text engines in June. Visit Leopard, Octopus, Porcupine, Rhino or Cobra platform pages and try web demos. Each demo, including speech-to-text, processes voice data within your web browser.
Ciao, こんにちは, 안녕하세요, Olá Polyglot Voice Interfaces
Picovoice introduced Japanese, Korean, Italian and Portuguese SDKs for Porcupine and Rhino in July. One can run eight speech models simultaneously and across platforms.
Voice Search in Eight Languages
Octopus Speech-to-Index might be just one year old, but within eight months, it started facilitating an accurate and efficient keyword search even for proper nouns in eight languages.
Community Updates
100 Hours of Monthly Transcription - for Free
Since the launch of the Free Tier last year, Picovoice has been the only voice AI vendor offering custom wake words, voice commands and voice activity detection that runs across platforms with no SDK or platform limit. Picovoice added 100 hours of free transcription and search to the Free Tier in March.
Alexa meets Porcupine
Seeing more developers can build with state-of-the-art speech models excites the humans of Picovoice. This month, it was not only Picovoice engineers enabling it. Federico Marotta wrapped Porcupine class to use with AVS (Alexa Voice Services). Don’t forget to share your projects with us via email or by tagging us on social media.
Most Popular
In case you missed any of them, we gathered the most popular tutorials, articles and SDK of 2022.
Tutorials
- How to Create Subtitles for any Video with Python
- Implementing a Lightweight Transcription Microservice with gRPC
- Speech-to-Text using JavaScript
Most Read Articles
- Tips for Choosing a Wake Word
- Local Speech-to-Text with Cloud-Level Accuracy
- Offline Voice AI in Web Browsers
Most Popular SDK
- Python was the developers’ first choice this year. If you haven’t tried, start building with a few lines of Python: