Demo Overview: Unlocking the Full Potential of Conversational AI with FPGAs | Achronix

The Achronix Automatic Speech Recognition (ASR) Demo converts spoken English to text up to 4K concurrent streams of audio processing when the performant demo is released in 2023. It’s 90% lower cost than a server-based solution. The solution can be hosted on-premises in your own data center which is ideal for highly sensitive speech analysis vs. a cloud-based solution. Achronix FPGAs accelerate AI performance and, unlike other solutions, can be reprogrammed with new/updated algorithms as needed. In this demo, we are showing an automatic speech recognition (ASR) solution running in our Speedster7t FPGA on a VectorPath card. This solution provides low latency real-time speech-to-text solutions that can reduce costs by 90% when using a single server and VectorPath accelerator card to replace 20 standard CPU-based server solutions that would be required to deliver the same performance. This solution is developed by our partner Myrtle.ai and can be deployed on standard, off-the-shelf hardware. This is a complete solution with a host API exposed so no RTL or FPGA knowledge is required. It is compatible with all major deep learning frameworks including PyTorch and Caffe2. Schedule a call with our experts today to discuss this application in your upcoming project: https://bit.ly/43n1fhv

Category