TTS API: Handling long text and natural-sounding pauses

Ben WalkerFeb 15, 2026

I'm using the TTS API to generate audiobook-style narration but struggling with:

1. Long texts getting cut off (seems to have a character limit) 2. No natural pauses between paragraphs 3. The voice sounds rushed on technical content

My current approach:

response = client.audio.speech.create(
    model="tts-1-hd",
    voice="onyx",
    input=long_text  # ~5000 chars
)

I've tried inserting "..." and newlines for pauses but results are inconsistent. Any tricks for more natural-sounding output?

3.2k views17 replies38 likesSolved

2 Replies

Jessica LiuStaffAccepted AnswerJun 17

Thanks for the detailed report. We've identified an issue with PDF parsing in the latest update. A fix is being deployed. In the meantime, try converting your PDFs to text format before uploading.

import pypdfdef pdf_to_text(pdf_path):
    reader = pypdf.PdfReader(pdf_path)
    return '\n'.join(page.extract_text() for page in reader.pages)

Sam Okwu Jun 18

The workaround works. Looking forward to the fix though, since we have 500+ PDFs and converting them all is painful.