Latest revision as of 10:03, 4 October 2026

Speakjet Phonetic Guide

The SpeakJet chip produces human speech and sound effects using its built-in **Mathematical Sound Architecture (MSA)**. Instead of playing back recorded audio files (like .WAV files), it uses math models to copy how human vocal cords, tongues, teeth, and lips move to make sound.

By sending single-byte numbers to the SpeakJet over a serial connection, you can string basic speech sounds together to form any word, sentence, or sound effect in real time.

How Speech Works on the SpeakJet

Human speech is made up of small sound building blocks:

• Phoneme: The general family of a sound (for example, the letter sound "T").
• Allophone: The exact version of that sound you actually hear, depending on where it sits in a word and what sounds come right before or after it.
  • The "T" sound at the start of tea has a sharp blast of air.
  • The "T" sound at the end of cat is shorter and softer.
  • The "T" sound in the middle of water sounds almost like a soft "D".

The SpeakJet has **72 built-in speech elements (allophones)**. As you send these sound codes to the chip, it smoothly glides the sound from one mouth shape to the next, just like a person talking.

Five Golden Rules for Clear Speech

Follow these five rules to make words sound clear and natural rather than muddy or robotic:

1. Spell by Sound, Not by Letters

English spelling does not match how words actually sound. For example, the letters "c-a-t" sound like "K-A-T", and "ph" sounds like "F". Always spell words using the sound codes that match what your ears hear, not how the word is written on paper.

2. Match Consonants to the Right Vowel

Consonant sounds (like B, D, G, K, S) change depending on whether the vowel that follows them is made in the **front** or the **back** of the mouth:

• Use \BE before front vowels (like Bear), but use \BO before back vowels (like Bone).
• Use \KE before front vowels (like Key), but use \KO before back vowels (like Comb).
• Use \SE before front vowels (like See), but use \SO before back vowels (like So).

3. Use the Right Start or End Sound

Some sounds need special versions depending on where they sit in a syllable:

• Starting a word/syllable: Use initial consonants like \BE, \DE, \GE, or \TU.
• Ending a word/syllable: Use ending consonants like \EB, \ED, \EG, or \TT.

4. Take care when Doubling Up Sounds Between Words

A few basic vowel sounds have a consistent sound from beginning to end and can be chained, but most other sounds can't be chained.

• If you say "Night Time", your mouth drops the "T" at the end of "Night" and blends directly into "Time" (saying "Ni-Time").
• If you program both "T" sounds back-to-back, the chip will stutter unnaturally.

5. Change the Pitch, Speed, and Bend

A voice that stays at the exact same pitch sounds flat and robotic.

• Questions: Raise the pitch code on the last word to make it sound like a question.
• Statements: Drop the pitch slightly at the very end of a sentence.
• Important Words: Slow down and raise the pitch slightly on the most important word in a sentence to add natural emphasis.

The Vowel Map

Vowels are made by holding your mouth open and changing where your tongue sits (High, Mid, or Low) and whether it is in the Front, Middle, or Back of your mouth. All back vowels are made with rounded lips.

Long Vowels vs. Sliding Vowels (Diphthongs)

• Making Pure Vowels Longer: To make a simple vowel sound longer, either send the same code twice in a row (for example: \IY \IY) or place the SLOW code (8) right in front of it.
• Sliding Vowels (Diphthongs): Sounds like \EYIY (the "A" in gate) or \AYWW (the "ow" in cow) are sliding sounds that move from one vowel to another. **Do not repeat these codes back-to-back**, or the chip will repeat the whole slide and sound like it is hiccuping.

Adding Brief Silences Before Pop Sounds

Sounds like P, B, T, D, K, and G need a tiny moment of silence before their release burst so the mouth can "build up air pressure."

• The chip adds a tiny gap automatically, but if a word sounds muddy, insert a short silence like Pause 4 (30 ms) or Pause 5 (60 ms) right before the starting pop sound.

Master MSA Command Guide

The MSA engine does not just play sounds—it has an entire programming language of control codes (decimal codes 0 to 31, and 255). These codes sit in the exact same 64-byte input buffer as your speech sounds and execute in the order they were sent.

Mastering these commands is what turns robotic, unintelligible babble into clear, expressive speech.

Single-Byte vs. Two-Byte Commands

• Single-Byte Commands: Codes like Fast (7), Slow (8), or Pause 2 (2) act on their own and only take up one byte in the buffer.
• Two-Byte Commands: Settings like Volume (20), Speed (21), Pitch (22), and Bend (23) **must be followed by an extra value byte (X)**. For example, setting the pitch to 90 requires sending two bytes in sequence: 22, 90.

Pauses and Silence Mechanics

The SpeakJet provides seven distinct pause commands (0 through 6). Picking the right pause is essential for clean sound transitions:

• Pause 0 (0 ms): Immediately tells the synthesizer to ramp its volume down to zero. Always put this at the very end of your speech data so the speaker doesn't hiss or pop when finished.
• Pauses 1, 2, and 3 (Soft Word Breaks):
  • Pause 1 = 100 ms
  • Pause 2 = 200 ms
  • Pause 3 = 700 ms
  • How they work: These pauses lower the volume while continuing to calculate the moving mouth shapes. Use these between words (Pause 1 or 2) and between whole sentences (Pause 3) for natural breathing pauses.
• Pauses 4, 5, and 6 (Hard Pop Silences):
  • Pause 4 = 30 ms
  • Pause 5 = 60 ms
  • Pause 6 = 90 ms
  • How they work: These pauses force the synthesizer to reach total, complete silence before changing the mouth shapes. Use these right before unvoiced pop sounds (like P, T, K) to clear out sound energy and make the consonant pop crisply.

Shaping Speed, Rhythm, and Timing

Real people do not pronounce every syllable at the same speed:

• Master Speed (Code 21, X): Sets the overall speaking rate. Default is 114 (range 0 to 127). Higher numbers make the voice talk faster.
• Fast (Code 7): Cuts the duration of the very next sound in half (0.5× length). Use this on small, unimportant syllables and helper words (like "the", "a", "of", "in").
• Slow (Code 8): Stretches the very next sound to 1.5× its normal length. Use this on stressed syllables or to draw out a vowel sound.
• Delay (Code 30, X): Freezes sound execution for X × 10 milliseconds (up to 2,550 ms or 2.55 seconds). Useful for timing gaps between phrases.

Vocal Muscle Tone: Stress and Relax

Human vocal cords and mouth muscles tighten when expressing emotion or emphasizing a syllable:

• Stress (Code 14): Simulates slightly tightening the mouth muscles on the next sound. This subtly shifts the vowel frequency upward. For example, playing Stress before \IH makes it sound closer to a sharp \IY.
• Relax (Code 15): Simulates loosening the mouth muscles on the next sound. For example, playing Relax before \IY softens it toward \IH.
• Rule for Doubled Sounds: Because Stress and Relax only apply to the single sound immediately following them, if you double a vowel (e.g., \EH \EH), you must send the command before both sounds: 14, 131, 14, 131.

Pitch and Emotion Modulation

Pitch represents the fundamental frequency (F0) of vocal cord vibrations:

• Pitch (Code 22, X): Accepts a value from 0 to 255 (default 88). The range covers 3 full musical octaves (32 Hz to 240 Hz).
  • Values below 30 vibrate so slowly that they sound like individual clicks instead of a tone.
  • Pitch code 0 completely turns off the vocal cord buzz (0 Hz).
  • Pitch changes only affect voiced sounds (vowels, nasals, voiced stops). They have no effect on whispered or unvoiced hissing sounds (like S, F, SH).
• Pitch Recipes:
  • Asking a Question: Keep your baseline pitch around 80–88, and on the very last syllable, step the pitch up: 22, 110.
  • Stating a Fact: Start at 90, and step down to 22, 75 on the final word to signal you are done talking.

Bend: Changing the Character of the Voice

• Bend (Code 23, X): Accepts values from 0 to 15 (default 5).
• Bend adjusts the frequency curves of the chip's internal oscillators, changing the perceived size and material of the vocal tract:
  • Low Bend (0 to 3): Makes the voice sound deep, hollow, and large.
  • Default Bend (5): Natural human voice balance.
  • High Bend (8 to 15): Shifts the resonance higher, making the voice sound small.

Flow Control, EEPROM Calling, and Port Pins

• Wait (Code 16): Pauses the chip's speech output until a Start command is received over the serial port or an input pin changes state.
• Repeat (Code 26, X): Repeats the very next code X times (0 to 255). Great for repeating sound effects or alarm pings without wasting buffer memory.
• Call Phrase (Code 28, X) and Goto Phrase (Code 29, X): Allows your code to trigger one of the 16 pre-recorded phrases stored inside the SpeakJet's EEPROM (phrases 0 to 15).
  • Call Phrase acts like a function call—it plays the phrase, and when it hits an End of Phrase code (255), it returns back to where it left off. You can nest calls up to 3 levels deep!
  • Goto Phrase jumps to the phrase and does not return.
• Port Control (Code 24, X) and Port Output (Code 25, X): Allows the SpeakJet's sound stream to directly flip external output pins (D0, D1, D2).
  • Send 24, 0 to take manual control of all three pins.
  • Send 25, Value to set them HIGH or LOW. You can use this to flash an LED on an exact syllable or ignite a model rocket after a countdown!
• Reset Defaults (Code 31): Instantly restores Volume (96), Speed (114), Pitch (88), and Bend (5) to factory default values.

Master Control Codes Summary Table

Allophone Speech Sound List (Codes 128–199)

This table lists all 72 basic speech building blocks built into the SpeakJet. Note that all lengths are approximate, and the actual measured lengths will vary based on the phoneme context in speech. (due to blending, volume fading, unvoiced phoneme stops, etc.)

Sound Effects & Tone List (Codes 200–254)

The SpeakJet also has 55 built-in sound effects, sirens, creature sounds, and telephone touch-tones.

Robot Chatter Sounds

Short, 80 ms electronic sounds useful for robotic chatter, button clicks, and sci-fi effects.

Sirens & Warning Alarms

Sustained pitch sweeps for alarms and alerts.

Beeps & Button Feedback

Biological & Animal Sounds

Telephone Keypad (DTMF) Tones

Standard dual-tone telephone keypad numbers. Each tone plays for 95 milliseconds.

Miscellaneous Sounds

Practical Code Examples

Example 1: Saying "Ready"

Break the word into four basic sounds:

1. Starting "R": \RR (148)
2. Short "e": \EH (131)
3. Front "D": \DE (174)
4. Ending "ee": \IY (128)
5. Finish with silence: Pause 0 (0)

// C code array to say "Ready"
const uint8_t SPK_READY[] = { 148, 131, 174, 128, 0 };

Example 2: Emphasizing a Word (Pitch + Slow + Stress)

In the sentence "Stop that!", you can make "Stop" urgent and loud by combining MSA commands:

1. Set Pitch higher: 22, 105
2. Slow down the 'O' vowel: 8, 135
3. Add vocal muscle tension: 14

const uint8_t SPK_STOP_THAT[] = {
    // Word 1: "STOP" (Emphasized)
    22, 105,             // Raise pitch to 105 Hz
    188,                 // \SO (S-sound before back vowel)
    191,                 // \TT (T-sound)
    8, 14, 135,          // Slow + Stress applied to \OH (Short O)
    199,                 // \PO (P-sound)
    
    4,                   // Pause 4 (30 ms gap between words)

    // Word 2: "THAT" (Lower pitch conclusion)
    22, 75,              // Lower pitch to 75 Hz
    169,                 // \DH (Voiced "TH")
    132,                 // \AY (Short "A")
    191,                 // \TT (Ending "T")

    0                    // Pause 0 (Clean finish)
};

Example 3: Asking a Question with Rising Pitch

Notice how raising the pitch on the final syllable of "Online?" makes it naturally sound like an inquiry:

const uint8_t SPK_IS_ONLINE[] = {
    // Baseline configuration
    31,                  // Reset to standard defaults (Pitch: 88 Hz)
    
    // "ON-" (\OH, \NO)
    135, 142,
    
    1,                   // Pause 1 (100 ms inter-syllable transition)

    // "-LINE?" (\LE, \OHIH, \NE) with rising inflection
    22, 115,             // Step pitch up from 88 Hz to 115 Hz
    145, 157, 141,

    0                    // Pause 0 (Fade to silence)
};

Example 4: "System Alert" with Horn and Tone Changes

Plays an alarm horn, speaks in a firm tone, pauses between words, and accents the final word:

const uint8_t SPK_SYSTEM_ALERT[] = {
    // 1. Reset standard sound settings
    31,                  // Reset defaults (Volume: 96, Speed: 114, Pitch: 88, Bend: 5)
    20, 110,             // Volume up to 110
    22, 70,              // Pitch down to 70 Hz (Lower, serious tone)

    // 2. Play Warning Horn
    215,                 // Alarm A5 (500 ms)
    5,                   // Pause 5 (60 ms silent gap)

    // 3. Speak "SYS-" (\SE, \IH, \SE)
    187, 129, 187,
    // "-TEM" (\TU, \EH, \MM)
    192, 131, 140,
    
    2,                   // Pause 2 (200 ms break between words)

    // 4. Raise pitch for "ALERT"
    22, 85,              // Pitch raised to 85 Hz for emphasis
    130,                 // "A-" (\EY)
    145, 14, 151, 191,   // "-LERT" (\LE, \Stress, \AXRR, \TT)

    // 5. Clean finish
    0                    // Pause 0 (Fade to silence)
};

Example 5: Dialing a Phone Number

Transmits the touch-tones for "555-1212" with brief pauses:

const uint8_t SPK_DIAL_SEQUENCE[] = {
    245, 6, 245, 6, 245, 2,      // 5, 5, 5 with small gaps and a 200 ms break
    241, 6, 242, 6, 241, 6, 242, // 1, 2, 1, 2
    0                            // Mute when done
};

Glossary of Terms

Allophone

A specific version of a basic speech sound. For example, the "T" sound in top has a strong burst of air, while the "T" sound in stop is softer. Both are allophones of the letter "T".

Bend

A setting (0–15) on the SpeakJet that changes the sound filters. Lower values make the voice sound deeper and hollow; higher values make it sound sharp and metallic.

Diphthong (Sliding Vowel)

A vowel sound that glides smoothly from one mouth shape to another within a single syllable, such as the "oy" in boy or the "ow" in cow.

Fricative (Hissing / Buzzing Sound)

A sound made by squeezing breath through a narrow opening in the teeth, lips, or tongue. Examples include "S", "Z", "F", "V", and "TH".

Nasal (Nose Sound)

A sound produced by letting air pass through the nose rather than the mouth, such as "M" and "N".

Phoneme

The basic building-block sound family in a language. Changing a phoneme changes the meaning of a word (like changing the "B" in bat to a "C" to make cat).

Pitch

How high or low a voice sounds, like musical notes on a keyboard. Pitch is measured in cycles per second (Hertz).

Pop Sound (Stop Consonant)

A sound made by completely blocking the airflow in your mouth for a split second and then releasing it with a burst, such as P, B, T, D, K, and G.

Resonant

A smooth, singing consonant sound that lets air flow freely through the mouth without hissing or buzzing, such as "L", "R", and "W".

Voiced Sound

A speech sound made while your vocal cords are vibrating. If you gently touch your throat while humming an "M", buzzing a "Z", or saying "AH", you can feel the vibration.

Voiceless (Unvoiced) Sound

A speech sound made purely with rushing air, without using your vocal cords. Examples include whispering, "S", "F", and "T".