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

I'm looking for someone with expert knowledge of Google Cloud Text-to-Speech, Chirp 3 HD, and the Agent Platform who can resolve a discrepancy between the public Python API and the Google Cloud Agent Platform Speech Playground. Using the same voice and the same visible synthesis settings, the Python API produces audio that is different from the Agent Platform. The differences include duration, pacing, pauses, loudness, waveform, and spectrogram. I have already verified that: Both use the same voice (th-TH-Chirp3-HD-Orus). Both use plain text input (not SSML). The synthesis settings are identical: LINEAR16 44,100 Hz Speaking rate = 1 Volume gain = 0 The Python script uses the public Google Cloud Text-to-Speech API. The Agent Platform sends requests through an internal Vertex AI / Agent Platform GraphQL service. I have attached: The Python script. A screenshot of the Agent Platform settings. A WAV generated by the Python API. A WAV generated by the Agent Platform. The objective is not to approximate the output. I am after a Python implementation to generate audio that is identical to the Agent Platform output. If reproducing the Agent Platform output through a Python script is not possible, I'm also open to alternative solutions that can automatically generate .wav files from a list of words in a .csv, provided the resulting audio is identical to what the Agent Platform would produce. The implementation language or method is not important - the priority is obtaining the exact same audio output automatically.