Google Cloud Text-to-Speech (Chirp 3 HD) API Script

Google Cloud Text-to-Speech seems to generate different audio through the Python API compared with Google Cloud’s online Agent Platform Speech Playground. I am using the same Thai Chirp 3 HD voice and apparently the same audio settings in both places, but the resulting WAV files differ in duration, pacing, pauses, loudness, waveform, and spectrogram. The goal is to determine exactly why the outputs differ and, ideally, modify the Python implementation so it reproduces the online Agent Platform exactly. The online Agent Platform interface shows: Task: Text-to-Speech Model: Chirp 3: HD Voices Language: Thai (Thailand) Voice: Orus Audio format: LINEAR16 Sample rate: 44100 Hz Speed: 1 Volume gain: 0 See attached screenshot. I have attached the Python script I am using. I have attached script_nǎng-sʉ̌ʉ-pim.wav and google_nǎng-sʉ̌ʉ-pim.wav to show that the generated files are different.