Overview
SMPGEN is a multi-engine AI sample generator, not a conventional sampler. Choose a generation backend, pick or write a prompt, then hit RENDER to create speech, music or other audio locally. Each output lands in the sample list on the right, where it can be edited and auditioned in the app.
The available engines include neural speech and audio generators, plus the classic Software Automatic Mouth (SAM) speech synthesizer. SAM is a traditional synthesizer, not an AI model. No cloud, no account and no model downloads are required at runtime: the AI model packs ship inside the roughly 2.8 GB Windows installer. See why the installer is big.
Draft manual, written against the v1.0 UI. The auto-generated manual (exported by the plugin itself via --export-manual) replaces this file on the next docs sync.
1. Generating audio
The GEN panel (top center) is the sample factory. Select a generation backend, enter a prompt and render the kind of audio that backend supports.
- PROMPTS list -- one click picks a line. Your own text goes in the field next to it.
- Generator selectors -- choose an available speech, music or audio engine.
Kokoro TTS (8 voices)is one neural speech option; select a model or voice variant where the engine offers one. SAM is the classic built-in speech synthesizer. - RENDER -- runs the selected generator and drops the finished take into the sample list.
- RND -- rolls a random prompt/variant; auto mode keeps generating variations hands-free.
- SPD -- generation speed for engines that support it, including speech generation.
Generation runs on CPU by default; the GPU toggle in the top bar switches the ONNX Runtime to the DirectML provider when a compatible GPU is present.
2. The sample list
Right-hand column, under RESCAN / FOLDER:
- Every rendered take is written to the SMPGEN sample folder and listed with a waveform preview, duration and format (e.g.
3.2s 44k st). - Click an entry to load it into the voice channel.
- RESCAN re-reads the folder after external changes; FOLDER opens it in the OS file manager.
- Your own WAV/AIFF files dropped in that folder show up here too.
3. Auditioning the generated sample
After rendering or selecting a take, use the bottom strip to audition and perform the generated audio:
| Control | What it does |
|---|---|
| Keyboard / OCT | On-screen keyboard + MIDI input; OCT buttons shift octaves. |
| TRIM | Set start/end on the waveform. |
| PLAY | Audition the loaded sample. |
| NORM | Normalize peak level. |
| REV | Reverse. |
| POLY | Polyphonic vs mono playback. |
| LOOP | Loop the trimmed region. |
| UNDO / REDO | Sample-edit history. |
4. The voice channel
Two rows of knobs between the generator and the keyboard shape the loaded sample per voice:
| Group | Controls | What it does |
|---|---|---|
| ADSR | ATK DCY SUS REL | Amplitude envelope (shown on the ADSR display). |
| FILTER | FLT DRV CRSH BITS | Filter amount, drive, bit-crush and bit depth. |
| EQ | LOW MID HIGH | Three-band tone EQ. |
| CHORUS | MIX RATE DPTH | Voice chorus. |
| MOD | PITCH VIB VRATE | Pitch offset, vibrato depth and rate. |
| DELAY | DLY FB MIX | Echo time, feedback, mix. |
| REVERB | MIX ROOM DAMP | Reverb amount, room size, damping. |
| OUT | SAT WID TONE PAN VOL | Saturation, stereo width, tone, pan, level. |
5. SMP / FX tabs
SMP -- the generator + voice channel described above.
FX -- the shared FX rack (200+ reorderable effects) plus the ffmpeg offline sample FX (trim, normalize, reverse and file-level processing).
6. LFO rack
The right-hand LFO RACK hosts unlimited LFOs (+ ADD LFO). Each LFO has a waveform, rate (free or BPM-synced divisions like 1/4), and AMP / SEED / PHASE knobs. Press LINK on an LFO, then touch any knob to bind the modulation.
7. Control panel
Left column, top to bottom:
| Section | Controls |
|---|---|
| I/O | GAIN input trim, VOLUME master out. |
| FILTER | Multimode MODE / CUTOFF / RESO on the summed output. |
| EQ | LOW / MID / HIGH. |
| COMP | THRESH / RATIO / KNEE compressor. |
| LIMIT | THRESH / RELEASE limiter. |
| PRESETS | Search box, SAVE, and ? (this manual). |
Top bar: THEME cycles the 16-color themes, GPU toggles the ONNX execution provider, ABOUT shows version/credits.
Why the installer is big
SMPGEN is heavier than the other AZOPSOFT plugins by design: the AI stack is inside the box, not behind a download.
| Piece | Why it ships |
|---|---|
| AI model packs: Kokoro, MusicGen, Piper, SA3Music, SAO Small, Supertonic and TangoFlux | Model weights for speech, music and other audio generation. |
| SAM (Software Automatic Mouth) | Classic speech synthesis engine, not an AI model. |
| ONNX Runtime + providers | Runs the models on CPU or DirectML GPU. |
espeak-ng-data | Phonemizer dictionary the TTS front-end needs for text to phonemes. |
| ffmpeg (static) | Offline sample FX (trim/norm/reverse/format work), no external install. |
The AI model payloads account for most of the installer size. Bundling them means the supported generators are available offline after installation, without separate model downloads.