NeuroSDR User Manual (2) — NeuroCaption and AF Plugins

NeuroSDR user manual (2) — NeuroCaption and AF plugins

Plugins attach to demodulated audio (or IQ) from your receiver. They are most powerful on local hardware with a stable sample stream; they also work on Virtual Kiwi/WebSDR when you use that secondary path. Author: KD8CEC.

1. Mental model

1. SET → AF Plugins → add a row → choose type → assign VFO (main or a Sub VFO) → APPLY.
2. Each SCENE owns its plugin instance list.
3. Decoder UIs appear as tabs on the right (FT8, CW, Caption, …).

YouTube: AF plugins & captions intro — 03:36

2. NeuroCaption engines

Engine	Use when	Notes
ONNX / Sherpa Whisper (Tiny/Base/Small)	Offline PC	Download INT8 model files; Base/Small cost more CPU than Tiny
Groq API	Fast cloud STT	API key in Caption SETUP
Self-hosted Whisper	Your GPU box	Host/port in Caption SETUP

YouTube: ONNX local — 04:07

YouTube: Translation (MyMemory) — 05:20

YouTube: Groq — 09:12

YouTube: Self-hosted Whisper — 12:16

YouTube: Tiny/Base/Small choice — 00:45

YouTube: Captions on AF waterfall ticker — 03:03

Caption quality follows SNR and program content. Clean SSB speech beats music beds on AM shortwave. Prefer Base/Small or Groq when Tiny invents text.

3. Shortwave schedule + AI scan

EiBi / shortwave scheduler style plugins browse broadcast schedules, filter by language/country, and jump the VFO. AI scan can look for live speech/language while you leave the radio on a hardware or remote HF path.

YouTube: Shortwave scheduler — 01:04

YouTube: AI language scan — 02:14

YouTube: On-air markers on the waterfall — 00:00

4. Decoder plugins (starter list)

Plugin	Role
FT8 / FT4	Weak-signal digital; multiple instances; MAIN WF overlay of callsigns
CEC CW	Multi-channel CW across the AF passband
SSTV / RTTY / Weather Fax	Classic HF utilities
flcw / flrtty / flfax	Fldigi-inspired AF decoders
Digital voice / DSD-FME	DMR, D-STAR, C4FM (advanced post)
AF Record	Plugin-scoped recording

YouTube: FT8 plugin setup — 01:04

YouTube: MAIN WF overlay — 06:33

5. RNNoise, AF bandpass, active CW filter

RNNoise extracts voice from noisy AM/SSB/FM audio — blend dry vs processed. AF bandpass cuts rumble/hiss. Active CW filter boosts Morse-shaped energy for ears or external decoders.

YouTube: RNNoise — 08:58

YouTube: Active CW filter — 11:33

YouTube: AF bandpass demo — 10:01

YouTube: RNNoise on RSP hardware — 07:00

6. Dual audio → other programs

SET → Audio → enable OUT2 → point at VB-CABLE → feed WSJT-X / FLDIGI / DeepCW while OUT1 stays on speakers.

YouTube: Dual audio + VB-CABLE — 03:12

YouTube: WSJT-X — 04:18

Next: Sub VFOs on wide hardware, satellite SCENE, digital voice, Smart Record, web remote.

---

NeuroSDR — KD8CEC · Blog: https://www.hamskey.com · Downloads: GitHub Releases · Source: github.com/phdlee/NeuroSDR