api.py generates normal audio with x86, but noise output with aarch64; infer-cli both fine

[G1017]

Checks

• This template is only for bug reports, usage problems go with 'Help Wanted'.
• I have thoroughly reviewed the project documentation but couldn't find information to solve my problem.
• I have searched for existing issues, including closed ones, and couldn't find a solution.
• I am using English to submit this issue to facilitate community communication.

Environment Details

#src/f5_tts/api.py
x86使用api.py生成正常的音频，aarch64使用api.py生成的音频是噪音
x86和aarch64使用src/f5_tts/infer/infer_cli.py都能生成正常的音频

Steps to Reproduce

#src/f5_tts/api.py
x86使用api.py生成正常的音频，aarch64使用api.py生成的音频是噪音
x86和aarch64使用src/f5_tts/infer/infer_cli.py都能生成正常的音频

✔️ Expected Behavior

No response

❌ Actual Behavior

No response

[G1017]

x86 uses api.py to generate normal audio, while aarch64 uses api.py to generate noise.
x86 and aarch64 use src/f5_tts/infer/infer_cli.py to generate normal audio.

[SWivid]

Have you debug on the input/output side, e.g. does the model get noise after torchaudio.load and before input into inference process, or all fine during inference but noise at end (output side)
And were ffmpeg all installed on both platform? Maybe also try replace

F5-TTS/src/f5_tts/infer/utils_infer.py

Line 401 in 52c8477

audio, sr = torchaudio.load(ref_audio)

with librosa loading, to see if torchaudio to blame