UB DSP Releases Vowel Blender - Phonetic Formant Filter With a 2D Vowel Pad and Tempo-Synced Motion Recorder
UB DSP has released Vowel Blender, a formant filter built around a 2D phonetic pad rather than the usual menu of preset vowel shapes. It is the developer's second plugin, following Grit Blender, and it treats vowel selection as something you play rather than something you dial in.
The central control is an XY pad mapped to ten vowels from the International Phonetic Alphabet (i, ɪ, ɛ, æ, ɑ, ʌ, ɔ, o, ʊ, u), positioned by tongue and jaw articulation. Drag the cursor and the formants morph between vowels in real time. The vowels are not approximations: each phoneme uses four formants modeled on real speech, drawn from two well-known phonetics datasets (Peterson & Barney 1952 and Hillenbrand et al. 1995). Seven voice models (three women, three men, and a kids model) shift the formant set to match a different vocal anatomy, so the same vowel path reads differently depending on the voice you pick.
What separates this from a static vowel effect is the motion recorder. You record a gesture on the pad and loop it in sync with the host, with forward, backward, and ping-pong playback, either free in Hz or synced from 1/8 to 8 bars. If you would rather not draw a path, there are 13 built-in shapes, from geometric loops like Triangle and Square to vocal paths like Diphthong and an all-vowels sweep. The effect works on any source, so a bass, synth, drum bus, or vocal can be made to speak, chant, or drift through less realistic formant territory.
Modulation and character controls
The modulation engine is deeper than the category usually offers. There are nine source types: LFO, MSEG, step sequencer, audio oscillator, envelope generator, envelope follower, macro, MIDI, and XY pad. Sources are multi-instance and drag-to-assign, and modulators can modulate other modulators, so a step sequencer routed through a macro through an LFO into the X axis is all valid signal flow.
A pre-formant pitch shifter (a Rust phase vocoder, ±12 semitones) pitches the source before the formant stage, keeping the vowel timbre grounded as the pitch moves; it adds around 12 ms of latency when active and no CPU when off. Three character knobs round out the sound: Humanize adds slow formant drift, Size scales the perceived vocal tract from child to giant, and Alien pushes the formants past anything a human voice would produce. A LOWS crossover lets the low end bypass formant processing entirely, useful on bass and full loops where you want the talking midrange without thinning the fundamental.
The filter itself is a ZDF SVF bandpass cascade at 24 dB/oct with gain compensation and a soft clipper, and the plugin ships with 38 factory presets across six categories.
Key features
- 2D phonetic XY pad with 10 IPA vowels, 4 formants each, morphing in real time.
- 7 voice models built from peer-reviewed phonetics research.
- Motion recorder with tempo-synced loop playback (forward, backward, ping-pong) and 13 built-in shapes.
- Modulation engine with 9 source types, multi-instance routing, and modulator-on-modulator intermodulation.
- Pre-formant pitch shifter (±12 semitones, ~12 ms latency, zero CPU when off).
- Humanize, Size, and Alien character controls, plus a LOWS crossover to preserve the low end.
- Stereo modes: Off, L/R, and M/S with adjustable stereo amount.
Pricing & Availability
Regular price $68. Introductory price $39 until July 1, 2026. Formats: CLAP, VST3, AU, AAX. Runs on macOS 10.13+ (Universal Binary, Intel and Apple Silicon), Windows 10+ (64-bit), and 64-bit Linux. A free unlimited trial is available, and purchases carry a 30-day money-back guarantee.
