1. What is the physical-modeling synthesizer Chimera
Chimera Vocal Synth is a physical-modeling voice synthesizer in the browser. It does not stitch pre-recorded voice clips or run a typical neural TTS pipeline. It simulates an airway: lung pressure, vibrating folds (or a creature syrinx), then a branched tract, lips, and nose.
The engine is Rust compiled to WebAssembly and runs in an AudioWorklet. Dragging the tongue or blending body size changes the acoustics themselves—not a different sample bank.
Think of it as:
- a playable virtual larynx + tract instrument;
- a sandbox for articulation and resonance;
- a creative tool for beatbox, melodies, and sound design.
This guide matches the web UI. Try it here: open the online synth.
2. Play mode: drag tongue and lips
Play is the best on-ramp: a head cutaway (or related views), articulator handles, and keys for sustained pitch while you listen to resonance change.
Before you start
- Turn system volume down.
- Click Start Audio or Breathe in and start. Without this, you usually hear nothing.
- Serve the page over HTTPS (or localhost). AudioWorklet needs a secure context.
Drag tongue, lips, and jaw
- Switch to Play.
- Hold a note (on-screen keys, computer keyboard, or MIDI).
- Drag tongue / tip / lips / jaw handles (labels vary by build).
- Move one articulator at a time.
- Try soft palate or larynx handles if shown.
Vowel pad and body blend
- Vowel pad / map: morph between vowels such as [i], [a], [u] and watch formants.
- Male / Female / Child: continuous body blend, not a recorded singer pack.
- Vibrato, Breathy, Ring, etc.—tweak after you can sustain a clean tone.
Computer keyboard
Many builds map letter keys (roughly A through semicolon) to notes, with Z / X for octave. Confirm against your embed; Web MIDI is often supported.
Exercise: vowel glide
Hold one pitch and slide the vowel pad from [i] toward [a], then [u]. That continuous change is the point of physical modeling.
3. Speak: free voice-synthesis style talking
Speak turns text into timed gestures and prosody that drive the same airway. Great for short phrases and phonetics—not a promise of commercial narration TTS.
English text
- Open Speak; leave Text is IPA off for normal spelling.
- Type something short like Hello from Chimera.
- Click Speak; Stop ends it; Rate changes speed.
- Ctrl/Cmd + Enter often works from the text box on desktop.
IPA
- Enable IPA for finer control.
- Use the consonant chart between vowels.
- Unsupported symbols are often listed; try near equivalents.
4. Sing: use Chimera as a browser singing tool
Sing maps lyric syllables onto pitch. New notes usually advance syllables; consonants often land before the beat.
- Try a built-in demo song if present.
- Enter Lyrics, then play keys—one new note, one syllable.
- Hold-through for legato; use hyphens to hold or split syllables.
- Without lyrics, keys often sing an “ah” vowel.
Lyric language support depends on the embedded build (e.g. English, Korean). Expression knobs matter more after a simple phrase works.
5. Beatbox
Pads and a step sequencer made of articulatory hits—not an 808 sample pack. Change one step at a time; keep early loops sparse.
6. Bestiary (animal voices)
Alternative physical vocal models—not animal sample libraries. Common creatures: songbird (two-voice syrinx), seal, alligator, elephant. Pick a creature and call; sustain with keys if available.
7. FAQ
Picture moves but no sound?
Confirm Start Audio / Breathe in, unmuted devices, HTTPS or localhost. Reload if the start overlay errors.
Is this commercial TTS?
Speak focuses on English spelling and IPA. Do not expect paste-any-language long-form narration.
Is Airway official?
No—third-party embed and guides. Contact 2811350864@qq.com.
8. Tips
- One variable at a time — articulators, body blend, breath, pitch.
- Short first — then lengthen Speak/Sing material.
- Use the meters — spectrum and pressure help map hearing to state.
- Keys + mouse — hold pitch while dragging the tract.
- External record — if there is no WAV export, capture with OS tools or OBS.
- Credit the engine when you share; do not pass off demos as yours.
Practice now
Path: Play sound → Speak short line → Sing two syllables → Beatbox two tracks → one Bestiary creature.