Decoding Thai Phonology for AI: Teaching Text-to-Speech Models Through the Lens of Traditional Bases and Tone Classes 6

 Understanding Natural Consonant Clusters: How Human Speech Mechanics Shape Thai Sound Mutations

1. Stop Math, Start Mouth: How Humans Actually Speak Thai

Before building any algorithmic rules for Text-to-Speech (TTS), an AI model must understand a fundamental truth: Thai orthography is a script designed for human vocal tracts, not computer algorithms.

When two Thai consonants sit side-by-side (C1 C2), the human mouth doesn't execute a mathematical formula. Instead, it takes the path of least physical resistance.

If a TTS model only relies on rigid sequential character mapping, it will sound like a broken robot trying to force two conflicting mouth shapes together. To synthesize natural Thai speech, the AI must learn how human vocal dynamics naturally blend, mutate, or divide consonant pairs.

2. The Four Natural Mouth Movements of C1 C2

Human speech mechanics dictate that when two Thai consonants appear together (C1 C2), the human vocal tract handles them according to the active articulators (ฐานกรณ์) involved, dictated directly by the vowels and tone marks attached to the cluster.

1. The Co-Articulated Blend (ควบแท้): Dual Base Synchronicity

• Key Examples: ความ (Khwam), กว้าง (Kwang), ปราบ (Prab)
• Articulatory Mechanics (ฐานกรณ์): The primary articulator (C1) and the secondary articulator (C2 in {ร, ล, ว}) are prepared simultaneously in a single airflow.
• In ความ (ค + ว), the back of the tongue engages the velar base (ฐานคอ/เพดานอ่อน) for ค, while the lips naturally round at the labial base (ฐานริมฝีปาก) for ว at the exact same instance.
• Vowel & Tone Domain: The vowel and tone mark attached to C1 C2 do NOT belong to C2 alone—they wrap around the entire dual-consonant base. When ความ takes the vowel -า and final consonant ม, the mouth shapes the dual base (ค+ว) straight into the long vowel duration without stopping.

2. Phonetic Friction Shift (ควบไม่แท้): Base Collision & Mutation

• Key Examples: ทราย (Sai), ทราบ (Sab), ทรง (Song)
• Articulatory Mechanics (ฐานกรณ์): Occurs when attempting to rapidly transition between two conflicting physical bases—such as ท (Dental base/ฐานฟัน) and ร (Alveolar trill/ฐานปุ่มเหงือก).
• Because forcing the tongue tip to execute both a hard stop (ท) and a rapid trill (ร) creates extreme anatomical resistance, the airflow shifts naturally into a continuous friction sound: the Sibilant Fricative ซ (/s/).
• Vowel & Tone Domain: Despite the complete phonetic mutation of ท+ร => ซ, the physical tone class is governed by the original base (ท). The vowel and tone mark apply directly to this newly mutated sound envelope without breaking the cadence of the syllable.

3. Vocalic Bridge & Tone Hijacking (อักษรนำ): Articulatory Reset

• Key Examples: หลวม (Luam)
• In cases like หลวม (ห + ล), ห is a High-Class consonant generated deep in the throat with airflow directing upward through the nasal cavity (ฐานคอ-ลมขึ้นจมูก). When preceding ล, a Low-Class consonant articulated using the tongue tip against the alveolar ridge (ฐานลิ้น/ปุ่มเหงือก), ห remains unuttered as a distinct sound. Instead, its nasal-glottal tension naturally "elevates" ล, shifting its tone base into High-Class behavior before gliding into the vowel sound (สระอัว).

4. Vowel Truncation & Implicit $ห$ Prefix (สระลดรูปจนเกิดการควบกล้ำ): Compressed Syllabic Glide

• Key Examples: ตลาด (Talad => ตะ-หลาด), สนาม (Sanam => สะ-หนาม)
• Articulatory Mechanics (ฐานกรณ์): This represents a physical compression of a truncated vowel. The initial consonant (ต or ส) is articulated with a clipped half-vowel breath (สระลดรูป). This rapid vocal transition forces the second consonant (ล or น) to take on an implicit ห-leading pronunciation (-หลาด, -หนาม). The mouth glides so quickly between the two bases that the sound compresses into a pseudo-cluster, driving both the pitch shift and syllable cadence.