Temporary presentation file (delete after use):rm frontend/src/app/phonetics-showcase/page.tsx

DomainCook Audio Synthesis Showcase

Comparative speech synthesis across ElevenLabs, AWS Polly, and Browser Web Speech based on manual input.

Architecture Features Built

Viterbi Dynamic Programming morpheme segmentation
Initialism dictionary (CRM, ERP, CX, JX, POS, AI, Bharat)
Bound suffix protection guard (cxcitement, techify)
Acoustic cadence anti-slur spacing (CRM crow vs Mcrow)
ElevenLabs flagship eleven_v4 model with Liam voice
AWS Polly neural SSML generator with spell-out tags & pauses
Cascading failover (ElevenLabs → AWS Polly → Browser Native)
90% usable quota safety threshold (10% hard reserve)
Presets:

5 Audio Output Variations:

1. ElevenLabs APIeleven_v4 • Liam
Spoken: “JX America CX”
2. AWS Polly (Without Decomposer)Raw Text
Input: “JX America CX”
3. AWS Polly (With Decomposer)Neural SSML
Spoken: “JX America CX”
4. Non-AI Browser (Without Decomposer)Web Speech API
Speaking: “JX America CX”
5. Non-AI Browser (With Decomposer)Cadence Enhanced
Speaking: “JX America CX”