Open-SSTV

Created by Kevin (W0AEZ) · Built with the assistance of Claude by Anthropic · ·

An open-source, cross-platform SSTV (Slow Scan Television) transceiver for amateur radio. Receives and decodes SSTV images live off your radio, and encodes and transmits images back, with optional Hamlib, direct serial, or TCI (ExpertSDR2 / SunSDR2 / AetherSDR) rig control.

Status: Beta (v0.6.9) — ready for user testing and feedback. TX and RX paths work end-to-end across all 22 supported modes, with a built-in QSO logbook (v0.4), an image gallery (v0.5), and opt-in remote control from a phone or laptop browser (v0.6). Rig control via rigctld or direct serial CAT is functional. Weak-signal decode is usable down to roughly 0 dB SNR on Robot 36.

Open-SSTV is looking for testers — on-air reports, audio captures of problem decodes, and UI feedback all welcome. Please file an issue with your OS, Python version, radio / interface, and any terminal output. For bug reports, open Settings → Diagnostics → Export Diagnostics… and attach the generated zip — it contains the recent log file, system info, and your redacted config, which is usually enough for a fix without follow-up questions. See Testing focus areas below for the spots we'd most like eyes on.

See CHANGELOG.md for the full release history. | User Guide

Goals

Open source end-to-end, GPL-3.0-or-later.
Cross-platform: Linux x86_64 and arm64, macOS (Apple Silicon; Intel via pipx), and Windows (experimental — prebuilt binaries available, real-hardware validation ongoing). Raspberry Pi 4/5 on-hardware testing planned.
Modern, intuitive UI built on Qt 6 (PySide6).
Lightweight enough to run on modest hardware. Pure Python + a small set of well-maintained scientific dependencies.
Real radio control via Hamlib's rigctld TCP daemon, direct serial (Icom CI-V, Kenwood/Elecraft, Yaesu CAT, DTR/RTS PTT), or TCI WebSocket (ExpertSDR2 / ExpertSDR3 / AetherSDR / SunSDR2) — so any supported radio works out of the box without an external daemon.
Decoder written from scratch because no maintained Python SSTV decoder exists on PyPI today. Algorithms mirror the well-known C reference slowrx.

Features

Transmit (TX)

22 SSTV modes -- Robot 36, Martin M1/M2/M3/M4, Scottie S1/S2/DX/S3/S4, PD-50/90/120/160/180/240/290, Wraase SC2-120/180, and Pasokon P3/P5/P7. See the Supported Modes table below.
Image editor -- crop, rotate, flip, and add text overlays (callsign, labels) before transmitting. Crop is locked to the target mode's aspect ratio. Text overlays support both named position presets (Top/Center/Bottom × Left/Center/Right) and pixel-precise X/Y spin boxes for fine placement.
QSO templates -- one-click text overlays for common QSO phases (CQ, Exchange, 73). Placeholder variables ({mycall}, {theircall}, {rst}, {date}, {time}) auto-fill from settings or prompt only for what's needed. Custom templates can be created, edited, and saved. Re-clicking a template auto-clears the previous text; a dedicated Clear Text button restores the clean image.
Layered template compositor (v0.3) -- richer per-mode templates built from photo, text, rectangle, gradient, RX-image, and station-image layers, composited at TX time onto the mode's native frame size. Live template gallery on the TX panel and a non-modal three-panel editor (Layers / live preview / Properties) with MMSSTV-style and named tokens (%c / {callsign}, %o / {tocall}, etc.).
+RX Image slot in the template editor (v0.3.1) for one-click reply layouts — composites the most-recently-received image into your outgoing TX frame. Default geometry sits in the bottom-right at 30%×25% with a small inset; resize and reposition like any other layer.
8 bundled fonts (v0.3.1) -- DejaVu Sans Bold, Inter Bold, Press Start 2P, Orbitron Bold, Oswald Bold, Exo 2 Bold, Bebas Neue, Share Tech Mono. Orbitron, Oswald, and Exo 2 ship as Google variable fonts and snap to their Bold weight axis automatically when the family name carries Bold intent. Drop-in user-supplied .ttf / .otf files in {user_config_dir}/open_sstv/fonts/ are still picked up automatically.
Rainbow gradient text mode (v0.3.1) -- per-text-layer Solid/Rainbow toggle. Rainbow paints a smooth HSV hue sweep through the glyph mask while preserving uniform stroke colour and the layer's alpha; horizontal text sweeps left-to- right, stacked text top-to-bottom.
Auto center-crop on TX (v0.3.2) -- before template compositing, the source photo is center-cropped to the mode's native aspect ratio and resized to the mode's frame size. Banner and overlay placement stays predictable for any input aspect (phone portrait, 4:3, 16:9, etc.) without distorting the photo. The original image is never modified; the manual image editor remains available for finer control.
Correct Robot 36 encoding -- custom line-pair encoder emits the canonical format that all real-world decoders (MMSSTV, SimpleSSTV, QSSTV, slowrx) expect. PySSTV's upstream Robot 36 produces a single-line format that most decoders cannot decode; Open-SSTV fixes this transparently.
PTT sequencing -- keys the rig, waits for a configurable relay-settle delay (0–2 s, default 200 ms), plays SSTV audio, then de-keys. Works with rigctld, direct serial CAT, DTR/RTS, or manual (VOX).
CW station ID -- optional Morse code callsign appended to every SSTV transmission (keyed under the same PTT). Satisfies the Part 97 identification requirement automatically. 15–30 WPM, 400–1200 Hz sidetone, 5 ms ramps to suppress key clicks. Test Tone is exempt (it's a calibration signal, not a communication). On by default; skipped silently with a warning if the callsign field is empty.
Test Tone -- a 700 Hz + 1900 Hz two-tone signal at −1 dBFS peak for 5 s, triggered from the Radio panel or the Audio Settings tab. Used for ALC calibration; the TX output gain slider remains live during playback so the operator can tune without stopping the tone.
TX output gain -- 0–100% slider with an optional "Enable overdrive" checkbox that expands the ceiling to 200% for setups where ALC won't move at 100%. Most USB-audio rigs (IC-7300, FT-991A, etc.) sit in the 10–15% range with overdrive off.
TX banner -- optional identification strip stamped across the top of every transmitted image. Shows your callsign flush-left and "Open-SSTV v{version}" flush-right. Three size presets (Small 24 px / Medium 32 px / Large 40 px) with matching font sizes, live preview in Settings, plus a "Preview on image…" button that shows the banner composited against a real photo before committing to TX. The source image is gently shrunk to fit below the strip so user content is never overwritten; output dimensions match the SSTV mode exactly. Configurable background and text colours; off by default.
Per-transmission TX watchdog -- two-stage timer bounds PTT exposure on a stuck encoder or hung audio driver. Stage 1 (encode) is a fixed 30 s budget; stage 2 (playback) is computed from the actual encoded sample count plus PTT delay with a 20 % margin and a 30 s floor, so a stuck Robot 36 aborts in under a minute while Pasokon P7 still gets its ~500 s.
Rig-swap lockout -- rig connect/disconnect controls are disabled for the full duration of a transmission so a mid-TX backend change cannot corrupt PTT state.
TX progress bar with elapsed/total time (at the active sample rate) and percentage.
Stop button -- abort mid-transmission; PTT is always de-keyed cleanly.
Export to Audio (v0.3.10) -- write a WAV file of the current TX panel composite (template + photo + QSO overlays + TX banner) without keying the radio. Uses exactly the same composited image Transmit would emit, so the WAV matches what would have gone over the air. Encode runs off the GUI thread; same job as open-sstv-encode but accessible without dropping to a shell. Disabled while live TX is in flight to prevent a mid-TX race.

Receive (RX)

Live decode -- start capturing from any audio input, and decoded images appear in a scrollable gallery strip as they arrive.
Per-line incremental decoder (default since v0.1.24) -- each scan line is decoded as soon as its sync pulse arrives rather than reprocessing the full growing audio buffer on every flush. O(1) work per line instead of O(N²) total, so the decoder stays ahead of real-time on Pi-class hardware even on the longest modes (Pasokon P7, Scottie DX). Covers all 22 modes including Robot 36's auto- detected per-line and line-pair wire formats. The legacy batch decoder is still available via a Settings toggle as a diagnostic fallback.
Progressive image preview -- the partial image updates line-by-line during reception so you can see it build in real time.
Optional final slant correction -- when enabled, a single-pass re-decode runs on the completed buffer with a global least-squares timing fit to compensate for TX/RX sound-card clock drift. Off by default: on weak or noisy signals the polyfit has no outlier rejection and can worsen the image. Robot 36 is explicitly skipped (different color pipeline between the incremental and batch paths).
Weak-signal mode -- optional relaxation of the VIS detection thresholds (leader presence 35 % → 20 %, start-bit minimum 17 ms → 15 ms) for signals audible in the noise that aren't triggering decode. False-positive VIS detections are handled gracefully (silent IDLE reset, no user-visible error).
Weak-signal robustness -- bandpass prefilter, median-filter click rejection, and adaptive rolling-threshold sync detection. Usable decode down to ~0 dB SNR on Robot 36; partial decode at −5 dB.
RX-during-TX gate -- the decoder is paused for the duration of every transmission, so the radio's own audio loopback never feeds back into the receive path. A 50 ms gate-off delay drains trailing RF before decode resumes; decoder state is reset between sessions to avoid stale residue.
Image gallery -- horizontal thumbnail strip of the 20 most-recent decodes, newest first, with context-menu Save-As / Copy-to-Clipboard actions. Images are persisted to a per-session temp directory to release PIL buffers from memory immediately after the thumbnail is rendered (in-memory fallback if temp-dir creation fails).
Auto-save -- optionally save every completed decode automatically to a configurable directory, with timestamped filenames (sstv_<mode>_YYYYMMDD_HHMMSS.png).
Save images -- manual Save button, Ctrl+S shortcut, gallery double-click, or right-click → Save As…
FFT waterfall (v0.3.5) -- floating spectrogram window (View → Waterfall) showing the 0–4 kHz SSTV audio band as a scrolling FFT, with distinct cool and warm palettes for RX and TX traffic and dotted reference lines at the 1200 / 1500 / 1900 / 2300 Hz SSTV tones. Hidden (not destroyed) when toggled off, so scroll history is preserved across openings; visibility persists across app restarts.
RX audio recording (v0.3.6) -- opt-in lossless capture of the raw received audio alongside each decoded image. Settings → Audio → "RX Audio Recording" toggles the feature and picks WAV (stdlib) or FLAC (~40 % smaller, requires the optional [flac] install extra). Lets operators re-decode marginal signals later via open-sstv-decode or a different decoder. Lossy formats are deliberately excluded because compression artefacts degrade re-decode quality.
Decode Audio (v0.3.10) -- offline-decode a .wav or .flac file from the RX panel; the result lands in the gallery exactly like a live decode. Pairs naturally with v0.3.6 RX audio recording for re-running marginal signals through the current decoder. Always enabled, including during live capture (results interleave in the same gallery).

Logbook & QSO Logging (v0.4)

Built-in QSO logbook -- every transmission and reception can be captured with its image, mode, frequency, UTC time, callsign, RSV report, and notes. Storage is a single schema-versioned SQLite file in the platform data dir; rows reference your image files, never copies of them.
Capture dialog -- opens pre-filled at TX/RX completion (TX drafts pull ToCall/RST/Name/Note straight from the QSO bar); Esc dismisses without writing anything. A config option logs silently instead.
Party-line aware -- SSTV calling frequencies carry everyone's QSOs, so an RX capture setting controls when receptions prompt: always, only while you're in a QSO (ToCall filled), or never -- and any decoded image can be logged after the fact from the gallery (right-click → Log QSO…).
Logbook window (Tools → Logbook…, Cmd/Ctrl+L) -- filterable table, image preview, edit/delete, manual entry with Save & New.
ADIF 3.1.5 import/export -- interop with HRD, N1MM+, LoTW (via TQSL), eQSL, Club Log, and QRZ.com. Exports stamp your station identity per record; imports dedupe on (callsign, time, mode). See docs/logbook.md for the full guide.
External Log over UDP (v0.6.7) -- an [External Log] button next to [Logbook…] on the TX panel's QSO bar broadcasts the current contact to companion logging software (QLog, JTAlert, GridTracker, Log4OM, N1MM…) the moment you're done working it, in raw-ADIF or WSJT-X's own framed protocol -- independent of the local logbook database.

Remote Access (v0.6)

Control from a browser -- an embedded, opt-in web server (off by default; loopback + token) lets a paired phone or laptop watch RX live, browse the gallery and logbook, and — behind an explicit gate and a connected CAT rig — compose and transmit. Enable it in Settings → Remote and pair by scanning the QR.
Compose & transmit -- take a photo (camera or upload), crop and frame it to the selected mode, add a station template, and send. Safety is layered: off by default, a single-writer control lease, per-transmit confirmation, and a dead-man's-switch that unkeys the rig if the browser goes silent.

Image Gallery (v0.5)

Built-in image browser -- browse received (and opt-in transmitted) images in a thumbnail grid, opened via Tools → Gallery… (Cmd/Ctrl+G). The gallery joins your image folder to the logbook: a logged image shows its contact's callsign / mode / frequency / time and a one-click → QSO jump; unlogged images still appear, dated and mode-tagged from the filename.
Filter & group -- by callsign, mode, or date range, sorted by date, callsign, or mode, with lazy on-demand thumbnails cached to disk so the grid stays smooth into the tens of thousands of images.
Operations -- re-send an image to the TX panel, export a copy, or delete the file (a linked logbook contact is kept, its image link cleared). See docs/gallery.md for the full guide.

Radio Control

rigctld (Hamlib) -- TCP client for rigctld, supporting PTT, frequency, mode, and S-meter. Auto-launch rigctld from the settings dialog.
Direct serial -- connect to your rig without an external daemon:
- Icom CI-V -- with preset picker for common models (IC-7300, IC-9700, etc.)
- Kenwood / Elecraft -- standard Kenwood command protocol
- Yaesu CAT -- FT-991/991A, FT-891, FT-710, FTDX10, FTDX101, FT-950, and FT-450/450D. The frequency command's digit width differs between the FT-450/450D (8 digits) and the newer rigs (9, zero-padded); it's auto-detected from the radio's own response, no per-model setting needed.
- PTT Only (DTR/RTS) -- simple serial PTT via DTR or RTS line
TCI (v0.3.5) -- WebSocket-based control for the Expert Electronics SunSDR2 family (ExpertSDR2 / ExpertSDR3) and the AetherSDR. A single ws://host:port connection (default 127.0.0.1:40001) carries both CAT control and binary PCM audio, so rig control and RX/TX audio share one transport with no virtual audio cables required.
FlexRadio direct -- control a 6000-series Flex over the SmartSDR TCP API with no rigctld and no virtual serial port: enter the radio's IP, pick a slice, and PTT / frequency / mode work directly. Audio still comes from your sound device (e.g. DAX). The S-meter is not available over this path (Flex streams meters as VITA-49/UDP) and reads as 0.
Band Plan (v0.3.6) -- one-click "Band Plan" popup button on the radio panel tunes the connected rig to any standard SSTV calling frequency. Twelve entries covering HF (80/40/20/17/15/10 m), VHF (2 m), and UHF (70 cm) with the correct mode (LSB below 10 MHz, USB above, FM on VHF/UHF). The 20 m 14.230 MHz USB primary is shown in bold. Button is disabled when no rig is connected or TX is in progress. A rejected frequency/mode change (VFO lock, band-edge, an unsupported CAT command) is now surfaced as a status-bar message instead of failing silently.
SSTV mode policy (Direct Serial only) -- Settings → Radio → Direct Serial → "SSTV mode" controls what the Band Plan button sends for the mode half of a tune, mirroring WSJT-X's rig Mode setting: Voice (default; sends the band-plan entry's plain USB/LSB/FM, unchanged from before), Data/Pkt (asks for the protocol's data-mode variant instead -- e.g. Yaesu DATA-U/DATA-L -- so SSTV doesn't land on plain USB with the speech processor still engaged; currently mapped for Yaesu CAT only, other protocols fall back to Voice), or Don't change mode (frequency only, for operators who already have their data mode set up manually).
Configurable baud rate -- 4800, 9600, 19200, 38400, 57600, or 115200 baud.
Rig status bar -- frequency, mode, and S-meter polled at 1 Hz when connected. Graceful disconnect: non-modal status bar message, auto-reconnect on next poll.

Settings & Configuration

Audio device selection -- separate input/output device pickers. RX input gain slider is 0–200 %; TX output gain slider is 0–100 % by default, expandable to 200 % with an "Enable overdrive" checkbox for setups where ALC won't move at 100 %. Device changes take effect immediately; a saved-but-missing device surfaces a status-bar notice on startup rather than silently falling back.
Cross-platform serial port enumeration -- uses serial.tools.list_ports for reliable port detection on Linux, macOS, and Windows. Port list is cached for 5 s so repeated Settings opens don't re-enumerate USB hardware.
TOML-based config -- all settings persist across sessions in a platform-appropriate config directory (~/.config/open_sstv/ on Linux, ~/Library/Application Support/open_sstv/ on macOS, %APPDATA%\open_sstv\ on Windows).
Resilient config loading -- malformed or missing config and template files fall back to built-in defaults instead of crashing. Legacy key names (e.g. pre-v0.1.24 experimental_incremental_decode) are migrated automatically.
RX no-progress watchdog timeout (v0.3.8) -- configurable via Settings → Receive (5–300 s, default 5 s). Bounds how long a decode can sit without emitting new lines before being aborted; higher values are useful for long QSB fades on weak signals, where walk_sync_grid's sync-bridge can resume once audio returns. Out-of-range values in hand-edited TOML are clamped on load.
Callsign -- saved in settings, pre-populated in the image editor's text overlay tool for quick QSO card creation, and used by both CW station ID and the TX banner.
Default TX mode -- pre-select your preferred mode so it is ready each session.

CLI Tools

open-sstv -- launch the Qt desktop application.
open-sstv-encode -- encode an image to an SSTV WAV file without the GUI.
open-sstv-decode -- decode an SSTV WAV file back into an image without the GUI.
Both CLI tools work without Qt installed, for headless or scripted use (Raspberry Pi, CI pipelines, batch processing).

Supported Modes

All 22 modes support both TX (encode) and RX (decode).

Mode	Resolution	Duration	Color System
Robot 36	320×240	~36 s	YCbCr
Martin M1	320×256	~114 s	RGB
Martin M2	160×256	~57 s	RGB
Martin M3	320×128	~57 s	RGB
Martin M4	160×128	~29 s	RGB
Scottie S1	320×256	~110 s	RGB
Scottie S2	160×256	~71 s	RGB
Scottie DX	320×256	~269 s	RGB
Scottie S3	320×128	~55 s	RGB
Scottie S4	160×128	~36 s	RGB
PD-50	320×256	~50 s	YCbCr
PD-90	320×256	~90 s	YCbCr
PD-120	640×496	~126 s	YCbCr
PD-160	512×400	~161 s	YCbCr
PD-180	640×496	~188 s	YCbCr
PD-240	640×496	~248 s	YCbCr
PD-290	800×616	~289 s	YCbCr
Wraase SC2-120	320×256	~122 s	RGB
Wraase SC2-180	320×256	~183 s	RGB
Pasokon P3	640×496	~203 s	RGB
Pasokon P5	640×496	~304 s	RGB
Pasokon P7	640×496	~406 s	RGB

Not yet implemented (need custom YCbCr 4:2:2 encoders not in PySSTV): Robot 8, Robot 12, Robot 24, Robot 72. Planned for a future release.

Screenshots

Rendered on Qt's cross-platform style rather than any one OS's window chrome, so they look like what Linux, macOS, and Windows users all get. Generated with scripts/capture_screenshots.py against synthetic contacts and testcards — no real log data.

Main window — TX panel with the template gallery filtered by role, the QSO bar (ToCall / RSTs / RSTr / Name / QTH / Grid / Note), mode selector with timing, and the RX panel waiting for a signal

Remote station (v0.6) — your station in a phone browser



Gallery — decodes stream in live over SSE. Take control* claims the single-operator lease; LIVE means the event stream is connected.*	Compose — shoot or upload a photo, frame it to the mode, pick a template, fill the QSO fields, transmit. Disabled until remote TX is enabled and* a CAT rig is connected.*	Logbook — the same contacts as the desktop log, each with its image and direction.

Logbook and gallery



Logbook (v0.4) — filterable QSO table, image preview with full contact details, ADIF import/export	Gallery (v0.5) — thumbnail grid joined to the logbook; filter by callsign, mode, or date, then re-send to TX

Capture dialog — pre-filled from the QSO bar at TX/RX completion, with the image that was sent or received	Image editor — crop, rotate, flip, and text overlays before encode

Settings



General — station identity, default TX mode, logbook capture behaviour, restore default templates	Audio — devices, gain, weak-signal mode, incremental decode

Radio — rigctld / Direct Serial / TCI / FlexRadio, PTT, and the SSTV mode policy for Band Plan tuning	Remote — the opt-in web server, LAN binding, pairing QR, and the separate remote-TX gate

Images — auto-save, TX banner with live preview, update-check opt-in	Logging — log level, log files, UDP QSO log destination, diagnostics export

Editors and dialogs



Template editor — Layers / Live preview / Properties, with sample QSO state for token previews	First launch — the one-time setup prompt for callsign and audio devices

About — version, the 22 supported modes, and licence

Architecture

PySSTV ──► encoder facade ──┐
   (Robot 36 uses custom    ├─► audio output ──► (radio TX via PTT)
    line-pair encoder)      │
                            │
       UI (Qt 6 / PySide6)──┤
                            │
       audio input ────────►├─► Decoder (FM demod -> VIS -> sync -> per-mode decode -> slant)
                            │       (pure NumPy/SciPy, no UI/IO deps)
       rigctld TCP ────────►│
       direct serial ──────►┤
       TCI WebSocket ──────►┘

The DSP core/ is a pure-Python package with no UI, audio, or socket dependencies -- it's unit-testable in headless CI and can be driven from a different front-end (TUI, web, CLI) without modification.

Install (prebuilt binary)

Prebuilt binaries for Windows, macOS (Apple Silicon), and Linux (x86_64 and arm64) are attached to every GitHub Release. No Python install required — just download, unzip, and run.

macOS (Apple Silicon)

As of v0.3.21 the macOS release ships as a real .app bundle (Open-SSTV.app), not a folder of files. Because it's ad-hoc signed (not yet Apple-notarized — planned), macOS Gatekeeper will refuse to launch it until you clear the quarantine flag that Safari/Chrome/Finder stamps on every downloaded file. You only need to do this once per download.

cd ~/Downloads
unzip -o open-sstv-macos-arm64.zip
xattr -cr Open-SSTV.app        # strip the quarantine flag from the whole bundle
open Open-SSTV.app             # or just double-click in Finder

After the one-time xattr -cr, double-clicking Open-SSTV.app in Finder launches it like any other Mac app — no Terminal step needed.

If the launch fails with library load disallowed by system policy, the quarantine flag wasn't fully stripped. Re-run xattr -cr Open-SSTV.app (note: the whole bundle must be the target, not just a file inside it).

Notarized builds are on the roadmap; when that lands, even the xattr -cr step goes away.

macOS (Intel) — no prebuilt binary

GitHub Actions retired the macos-13 (Intel) runner pool, which is the last x86_64 macOS image they hosted; Apple-Silicon-only is now the only practical CI option for macOS releases. Universal2 is not viable for us either: PySide6 ships per-architecture wheels rather than a Universal2 fat binary, so a target_arch="universal2" PyInstaller build fails at collect time.

If you're on an Intel Mac, install from source instead:

python3 -m pip install --user pipx
pipx install open-sstv          # or: pipx install "open-sstv[flac]"
open-sstv

pipx puts Open-SSTV in an isolated venv on PATH and pulls the appropriate x86_64 wheels from PyPI for every native dependency. This is the same install path the development setup below uses, just without the clone.

Linux (x86_64 or arm64)

Two formats are published per architecture:

.AppImage — single-file, self-contained. chmod +x open-sstv-*.AppImage && ./open-sstv-*.AppImage.
.zip — unpacked onedir bundle. unzip open-sstv-linux-*.zip && ./open-sstv/open-sstv.

Windows

Download open-sstv-windows.zip, unzip, and double-click open-sstv.exe. Windows SmartScreen may show a "Windows protected your PC" dialog on first launch because the binary is unsigned; click More info → Run anyway.

Install (development)

Linux and macOS (supported)

git clone https://github.com/bucknova/Open-SSTV.git
cd Open-SSTV
python -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

You will also need Hamlib's rigctld for rigctld-based radio control (not required for direct serial or manual PTT):

macOS: brew install hamlib
Debian/Ubuntu: sudo apt install libhamlib-utils

Windows (experimental — untested)

⚠️ Open-SSTV has not been tested on Windows. Every runtime dependency (PySide6, numpy, scipy, sounddevice, PySSTV, Pillow, pyserial) publishes Windows wheels, so the app should install and run, but no one has yet driven a real radio from Open-SSTV on Windows. Please treat the instructions below as a call for testers rather than a supported install, and file issues with your findings — good or bad. Full validated Windows support is on the v0.3 roadmap.

Prerequisites: Python 3.11+ from python.org (tick "Add Python to PATH" during install) and a working Git for Windows.

From a cmd.exe or PowerShell prompt:

git clone https://github.com/bucknova/Open-SSTV.git
cd Open-SSTV
python -m venv .venv
.venv\Scripts\activate
pip install -e ".[dev]"

For rigctld-based radio control, download Hamlib for Windows from the official releases (pick a hamlib-w64-*.zip), unzip somewhere permanent, and either add its bin\ folder to PATH or launch rigctld.exe manually before starting Open-SSTV. Direct serial rig control (Icom CI-V, Kenwood, Yaesu, DTR/RTS PTT) does not require Hamlib.

Known Windows caveats (expected, not yet verified on hardware):

Serial ports appear as COM3, COM4, etc. — no /dev/... paths. serial.tools.list_ports enumerates them natively, so the Settings dialog's port picker should populate automatically.
Audio devices — use the MME or WASAPI host API. ASIO is not exposed by sounddevice out of the box. If your USB interface doesn't appear, open it once in the Windows Sound control panel to register the endpoint.
Config directory — settings persist to %APPDATA%\open_sstv\ (via platformdirs), not the Linux/macOS paths mentioned elsewhere in this README.
rigctld auto-launch from the Settings dialog assumes rigctld is on PATH; if not, launch it manually and connect Open-SSTV to 127.0.0.1:4532 instead.

Run

open-sstv                                              # Qt desktop app
open-sstv-encode in.png --mode martin_m1 -o out.wav   # CLI encoder
open-sstv-decode in.wav -o out.png                    # CLI decoder

Testing focus areas

If you're kicking the tyres on the v0.6 beta, these are the surfaces we'd most like eyes on. File an issue with what you tried and what happened.

Remote station (v0.6) — the newest and least-proven path, and the one that can key your transmitter. Pair a phone, watch a live decode, then try composing and sending into a dummy load first. Confirm the dead-man's-switch: start a transmission, then close the tab or walk out of Wi-Fi range, and check the rig unkeys. Reports on phones/browsers we haven't tried are especially useful.
Logbook and gallery (v0.4 / v0.5) — capture a QSO from a decode, export ADIF into your usual logger, and check the gallery↔logbook cross-links land on the right contact.
Weak-signal RX. Decode quality on fading / noisy signals; the weak-signal mode toggle (Settings → Audio → Receive); false-positive VIS detections (expected to reset silently — report if they don't).
RX decoder watchdog (v0.1.36). Intentionally interrupt a transmission mid-image — does the partial image land in the gallery? Does the decoder return to IDLE cleanly for the next VIS?
TX output level calibration. Test Tone button, output-gain slider, overdrive toggle. Does ALC move predictably? Any odd interactions with your rig's USB MOD Level or the OS system volume?
CW station ID. At 15 / 20 / 30 WPM, with your real callsign. Audible and legible to a human copying by ear?
TX preview outline (v0.1.37). Load an image, walk through the mode dropdown — does the green/amber match indicator track what you'd expect?
Rig control edge cases. Mid-session USB unplug; rigctld daemon crash; Icom CI-V addresses other than the default 0x94; Kenwood/Yaesu protocol quirks.
TCI rigs (v0.3.5). If you have an ExpertSDR2/3, AetherSDR, or another TCI-speaking SDR, on-air reports are especially valuable — this path is newly added and has only been validated against one AetherSDR setup. Confirm TCI connect, RX audio routing, full SSTV TX, and CW ID over TCI.
FFT waterfall (v0.3.5). Toggle View → Waterfall and confirm RX traffic paints a cool palette and TX audio paints a warm palette during a transmission; report any visible scrolling stalls or palette glitches.
macOS privacy prompts. If you see Music / iCloud / unexpected access requests on launch, note which ones and when — we have a hunch this is PortAudio device enumeration but haven't nailed it yet.
Windows. Now in real use, and the CI matrix builds and tests on Windows, but it sees far less on-air mileage than macOS and Linux — audio device quirks (MME vs WASAPI) and rig-control paths are where problems have surfaced so far. Reports very welcome.

Roadmap

Remaining SSTV modes -- Robot 8/12/24/72 (4 modes needing custom YCbCr 4:2:2 encoders not yet in PySSTV).
Raspberry Pi 4/5 validation -- Linux arm64 binaries already ship; on-hardware testing still pending.
Digital VOX -- auto-detect incoming SSTV and start decoding without manual capture start.
Drag-and-drop image loading in the TX panel.

Future

Expanded template library -- a full set of premade QSO templates inspired by MMSSTV and other popular SSTV clients (signal reports, contest exchanges, ragchew layouts).
Expanded font support -- more typeface options for text overlays, including styles common in amateur radio use.
Advanced text layout -- multi-column overlays, alignment controls, and background fill options for finer control over callsign and caption placement.
FSKID transmission (CW ID already shipping — see Features → Transmit).
LoTW (TQSL), eQSL, and QRZ.com direct upload from the logbook (ADIF export covers them today — see Features → Logbook).
PSK Reporter / DX cluster spotting.
Installer packaging (.deb, .dmg, Flatpak).
PyPI publish.
Plugin/macro system.
Internationalization.

Considered and deferred

Optional image post-processing — non-destructive "clean up image" action in the gallery (median filter for salt-and-pepper noise). Deferred: simple filters trade detail for smoother noise; more advanced denoisers (NLM, neural) bring significant dependency weight for mixed returns. May revisit if a concrete use case emerges.

Author

Open-SSTV is developed by Kevin (W0AEZ).

License

GPL-3.0-or-later. See LICENSE.

Name		Name	Last commit message	Last commit date
.github/workflows		.github/workflows
assets		assets
design/remote		design/remote
docs		docs
packaging		packaging
scripts		scripts
src/open_sstv		src/open_sstv
tests		tests
.gitignore		.gitignore
CHANGELOG.md		CHANGELOG.md
LICENSE		LICENSE
README.md		README.md
SSTV_App_User_Guide.md		SSTV_App_User_Guide.md
open_sstv.spec		open_sstv.spec
pyproject.toml		pyproject.toml

Latest commit

History

Folders and files

Repository files navigation