Changelog - All Posts

🤖 Models​

• Introducing Digital/v1 and Phone/v1: We are proud to announce our next-generation models, Digital/v1 and Phone/v1. These models deliver enhanced performance with improved accuracy and reduced latency compared to their predecessors. For comprehensive technical specifications and benchmarks, please visit our model index page.
• Legacy Model Deprecation: The Telephone/v6 and Default/v10 models are now entering deprecation phase. While they remain accessible on the platform, we strongly encourage migration to the new Digital/v1 and Phone/v1 models for optimal performance.

✨ Features​

• Real-time Streaming Support: We have launched a new WebSocket endpoint enabling real-time audio verification capabilities. For implementation details and best practices, please refer to our streaming documentation.
• Enhanced Console Analytics: The Console's Overview page now includes a streamlined dashboard for monitoring monthly credit usage, providing better visibility into your resource utilization.

🛠️ Fixes​

• Chrome extension signup: We have solved an issue that prevented users from properly verifying their emails while using the Chrome Extension.

🤖 Models​

• Telephone model: We are releasing our new Telephone/V6 model. This updated model will provide more accurate results and enhanced coverage through the implementation of a wider set of audio codecs.
• Parameter adjustment: We are adjusting the minimum voice duration parameter for audio chunks in authenticity verifications. This will provide more robust results in audios of conversational use cases.

✨ Features​

• Credit system: Users can now conveniently check their available credits within the console interface. Each credit allows for the verification of one minute of audio.
• Web Demo: our web demo now allows for both batch uploads as well as provides a time-based view of scores when a single file is uploaded.
• Console: A new timezone selection option has been added to the console filter when checking records and resources, enabling users to select the desired timezone for checking their verifications and audios.

🛠️ Fixes​

• API messaging: We have improved our API message responses to provide clearer information in cases of insufficient voice duration.

💥 Changes​

• Voice duration calculation has been optimized to occur during verification processing rather than at upload time.

🤖 Models​

• Introducing our latest v10 default model and v5 telephone model for authenticity verifications, delivering superior accuracy especially for audio containing background music.

✨ Features​

• Implemented secure password reset functionality through the Console interface
• Enhanced regional flexibility with EU/US region selection in both Console and Demo environments
• Expanded filtering capabilities with new verification type filter
• Introduced granular verification score range filtering in the Console

🛠️ Fixes​

• Improved error handling with corrected HTTP 5XX status code mapping

🤖 Models​

• Enhanced and optimized the v9 default model for authenticity verifications, providing more robust results against background music.

✨ Features​

• Introduced a Performed Without Score state to verifications, improving result readability. This state is now visible in the Platform Console.
• Added Chunk type explanations in the Platform Console, accessible within the Chunks view of a verification.

🛠️ Fixes​

• Resolved inconsistent behavior in audio files with high background noise levels.
• Optimized the search algorithm within the Platform Console.
• Addressed improper HTTP 500 errors caused by corrupted audio files.

🤖 Models​

• Introduced a new default model for authenticity verifications, offering improved resilience against background music.

✨ Features​

• Implemented enhanced voice activity detection and processing, improving overall system reliability.
• Developed Chrome extension functionality, preparing for the upcoming release of our Deepfake Voice Detection extension.
• Integrated search functionality in the web console.

🛠️ Fixes​

• Corrected error mapping to properly distinguish between HTTP 400 and HTTP 500 errors.

🤖 Models​

• Launched a new v9 default model for authenticity verifications, delivering enhanced robustness against background music.

✨ Features​

• Released the Hiya Deepfake Voice Detector Chrome Extension, enabling audio verification directly through your browser.

🛠️ Fixes​

• Resolved signup redirection issues in the web console.
• Corrected email notification system for user account verification.
• Fixed Slack integration links for user creation verification.

No posts yet