ai-coustics vs DPDFNet

ai-coustics provides audio-processing tools for Voice AI systems, preparing incoming sound for speech recognition and other downstream tasks. Its Quail models enhance speech and can reduce word error rate by as much as 30%; Quail Voice Focus suppresses competing voices to isolate the foreground speaker. Tyto predicts whether audio may cause downstream failures and returns a score with the cause. A voice activity detection model is designed for complex real-world environments without a separate denoising tool. The SDK runs enhancement, voice activity detection, and audio analysis on CPU within the customer’s application. The company says on-device processing has sub-40 ms latency, and audio processed by the SDK remains on the customer’s infrastructure. Integration options include Python, Node.js, C, C++, Rust, and WebAssembly, with guidance for LiveKit and Pipecat. A free SDK trial is offered through the Developer Platform, and paid plans start at $135 a month billed annually. Startup includes 100,000 minutes per month; Pro is $360 a month billed annually, Business is $540 a month billed annually, and Enterprise starts at $2,000 a month. The company’s own infrastructure runs in AWS eu-central-1 in Frankfurt, Germany.

DPDFNet is a free, open-source model for reducing noise in single-channel speech from recordings, live streams, and edge-device applications. It supports offline processing and streaming, and extends DeepFilterNet2 with dual-path encoder blocks and an always-on fine-tuning approach intended to reduce over-attenuation. Pretrained variants support 8, 16, and 48 kHz audio. The repository includes PyTorch code and ONNX and TFLite inference models, with command-line, Python, and stateful streaming interfaces. Streaming buffers about 20 ms before the first enhanced output, then adds about 10 ms of delay for later blocks. A hosted Gradio demo and audio examples are linked from the project page. Listed community integrations include sherpa-onnx, OBS Studio, VST3 and LADSPA plugins, Core ML ports, and WebAssembly. The project page reports real-time DPDFNet-4 deployment on Ceva-NeuPro-Nano edge NPUs in a benchmark setup using int8 weights and int16 activations. The software is under Apache License 2.0. Model files are not bundled in the source repository; it provides commands to download checkpoints and ONNX or TFLite models from Hugging Face.