RADE is an HF digital-voice approach that combines machine learning with classical DSP to deliver high-quality speech in 1500 Hz of RF bandwidth, with reported operation down to −2 dB SNR. FreeDV integrates RADE V1 into the FreeDV-GUI application (v2.0 and later), enabling practical over-the-air use with typical PC resources.
Primary link: https://freedv.org/radio-autoencoder/
ChatRF is a modular, Python-based repeater controller that embeds a conversational AI assistant into normal repeater operations. It supports interaction via DTMF, automatic CW identification, and optional features like weather, satellite tracking, and callsign lookups while using a local LLM (default: gemma3 via Ollama) for responses.
Primary link: https://github.com/zisispolychronidis/ChatRF/
This SDR-centric system captures repeater audio, transcribes it with the Whisper speech-to-text model, and produces timestamped logs for monitoring and review. The repository describes an RTL-SDR front end, Flask-based web/API components, and optional callsign validation via QRZ.
Primary link: https://github.com/Wintergrasped/KK7NQN-TranscriptionLogger
RadioTranscriptor is an open-source Python tool that performs real-time transcription of SDR audio using Whisper, typically paired with voice-activity detection and session logging. The project notes optional GPU acceleration (CUDA when available) and recommends 48 kHz audio input that is resampled for Whisper compatibility.
Primary link: https://github.com/theckid/RadioTranscriptor
This documented prototype uses an SDR demodulator to convert RF to audio, applies speech-to-text, and then uses sequence alignment against a callsign database to identify likely callsigns and messages. It is useful as a reference architecture for voice-net logging or assisted operator training experiments.
Primary link: https://medium.com/data-science/make-amateur-radio-cool-again-said-mr-artificial-intelligence-36cb32978fb2
DeepFilterNet provides deep-learning-based noise suppression that can be applied to recorded or routed receiver audio as a post-processing or near-real-time filter. Amateur-radio commentary has specifically pointed to the concept of wrapping DeepFilterNet into an operator-friendly, real-time “AI filter” for HF voice reception.
Primary link: https://github.com/Rikorose/DeepFilterNet
A real-time Morse (CW) decoder that runs entirely in the browser, using a CRNN model trained on programmatically generated Morse audio and executed via ONNX Runtime Web. This architecture enables a “no install” workflow while still leveraging modern ML inference on the client side.
Primary link: https://github.com/e04/web-deep-cw-decoder
morseangel is a Python application that decodes Morse code directly from an audio device using a deep neural network implemented in PyTorch, with a PyQt GUI. It is well-suited for experimentation with model-based decoding compared with rule-based CW decoders.
Primary link: https://github.com/f4exb/morseangel
Created by AG1LE, deepmorse-decoder is a deep-learning Morse decoder whose repository documents a CNN-LSTM-CTC model and includes retraining capability. It can be used as a practical starting point for building custom CW decoders tailored to specific band conditions, audio chains, or sending styles.
Primary link: https://github.com/ag1le/deepmorse-decoder
gr-dnn is an open-source GNU Radio out-of-tree module that runs deep-learning inference within GNU Radio flowgraphs using an ONNX-based inference engine. The associated GNU Radio Conference paper demonstrates inference on raw I/Q samples acquired with an SDR (e.g., PlutoSDR), enabling ML-enabled signal classification pipelines.
Primary link: https://pubs.gnuradio.org/index.php/grcon/article/view/69
This build-oriented project performs real-time RF modulation recognition by running CNN inference on an FPGA-based DPU, with an RTL-SDR as the receiver front end. The author explicitly frames the workflow around monitoring “random RF radio signals” (including examples on the 2 m amateur band) and classifying modulation in real time.
Primary link: https://www.hackster.io/matjaz4/rf-modulation-recognition-with-gnu-radio-11b294
This GNU Radio Conference paper motivates and documents public synthetic datasets for radio machine learning, emphasizing the need for common benchmarks in RF ML. For amateurs building modulation classifiers, signal identifiers, or band-activity analytics, these datasets and the associated generation approach are a standard reference point.
Primary link: https://pubs.gnuradio.org/index.php/grcon/article/view/11
Chris. “Make Amateur Radio Cool Again”, Said Mr Artificial Intelligence.” Medium, 6 Oct. 2018, https://medium.com/data-science/make-amateur-radio-cool-again-said-mr-artificial-intelligence-36cb32978fb2. Accessed 11 Jan. 2026.
e04. “web-deep-cw-decoder: A CW (Morse Code) Decoder Powered by Deep Learning.” GitHub, n.d., https://github.com/e04/web-deep-cw-decoder. Accessed 11 Jan. 2026.
f4exb. “morseangel: Deep Neural Network for Morse Decoding.” GitHub, n.d., https://github.com/f4exb/morseangel. Accessed 11 Jan. 2026.
FreeDV. “RADE – Radio Autoencoder.” FreeDV, n.d., https://freedv.org/radio-autoencoder/. Accessed 11 Jan. 2026.
O’Shea, Timothy J., and Nathan West. “Radio Machine Learning Dataset Generation with GNU Radio.” Proceedings of the GNU Radio Conference, vol. 1, no. 1, 6 Sept. 2016, https://pubs.gnuradio.org/index.php/grcon/article/view/11. Accessed 11 Jan. 2026.
Rikorose. “DeepFilterNet: Noise Suppression Using Deep Filtering.” GitHub, n.d., https://github.com/Rikorose/DeepFilterNet. Accessed 11 Jan. 2026.
Rodriguez, Oscar, and Alberto Dassatti. “Deep Learning Inference in GNU Radio with ONNX.” Proceedings of the GNU Radio Conference, vol. 5, no. 1, 12 Sept. 2020, https://pubs.gnuradio.org/index.php/grcon/article/view/69. Accessed 11 Jan. 2026.
Romanchik, Dan (KB6NU). “Artificial Intelligence and Machine Learning for Amateur Radio.” KB6NU’s Ham Radio Blog, 2 Dec. 2022, https://www.kb6nu.com/artificial-intelligence-and-machine-learning-for-amateur-radio/. Accessed 11 Jan. 2026.
theckid. “RadioTranscriptor: A Voice Transcriptor for SDR.” GitHub, n.d., https://github.com/theckid/RadioTranscriptor. Accessed 11 Jan. 2026.
Wintergrasped. “KK7NQN-TranscriptionLogger: Transcribe and Log HAM Radio Traffic.” GitHub, n.d., https://github.com/Wintergrasped/KK7NQN-TranscriptionLogger. Accessed 11 Jan. 2026.
Zibert, Matjaz. “RF Modulation Recognition with GNU Radio.” Hackster.io, 18 Apr. 2022, https://www.hackster.io/matjaz4/rf-modulation-recognition-with-gnu-radio-11b294. Accessed 11 Jan. 2026.
zisispolychronidis. “ChatRF: AI-Enhanced Ham Radio Repeater.” GitHub, n.d., https://github.com/zisispolychronidis/ChatRF/. Accessed 11 Jan. 2026.
Niininen, Mauri (AG1LE). “deepmorse-decoder: Deep Learning Morse Decoder.” GitHub, n.d., https://github.com/ag1le/deepmorse-decoder. Accessed 11 Jan. 2026.