Twilio Play 40-minute file limit vs Sume TTS 1,200-second cap

Twilio's <Play> documentation lists mp3, wav, aiff, gsm and ulaw as supported formats and warns that playing a file longer than 40 minutes can drop the call. Sume TTS fails any synthesized audio over 1,200 seconds (20 minutes) with tts_duration_exceeded, so a single Sume prompt can never hit Twilio's ceiling. Use mp3 for convenience or pcm_mulaw at 8,000 Hz for a phone-native file.

Twilio's page versus Sume's output

Twilio's page also recommends lossless 8 kbps formats over 128 kbps mp3 for voice, and notes that caching depends on ETag and Last-Modified headers.

A phone-ready request

Ask for wav, pcm_mulaw and an 8,000 Hz sample rate. Host the artifact at a public HTTPS URL and point <Play> at it.

import os, requests

r = requests.post(
    "https://api.sume.com/v1/tts-1.0/generate",
    headers={"Authorization": f"Bearer {os.environ['SUME_API_KEY']}",
             "Idempotency-Key": "hold-msg-001"},
    json={"transcript": "Thanks for holding. An agent will be with you soon.",
          "avatar_id": os.environ["SUME_AVATAR_ID"],
          "output_format": {"container": "wav", "encoding": "pcm_mulaw",
                            "sample_rate": 8000}},
)
print(r.status_code, r.json())

Check before launch

Test the file on a real call. Whether Twilio treats your hosted URL as cacheable depends on the headers your host returns, which Twilio's page says matter.

Worked example

Think about the two ways this fails.

• Long file: a 45-minute recording will drop the call on Twilio even if the file is valid; Sume TTS refuses to produce more than 1,200 seconds anyway.
• Wrong format: an mp3 plays on Twilio, but Twilio's page recommends lossless 8 kbps formats for voice, which is why pcm_mulaw at 8,000 Hz is offered.
• Cache surprises: Twilio's page says caching depends on ETag and Last-Modified.

Checklist before you commit

Host the file where you control headers if your prompt changes often.

• Test the prompt on a real handset.
• Keep hold messages short.
• Regenerate with a new filename when the text changes.