Deepgram Aura-2 allows 2,000 characters per request; Sume 20,000

Deepgram's text-to-speech guide says Aura-2 and Aura-1 take at most 2,000 characters per request, and that a longer payload "can result in a 413: Input Text Exceeds Character Limits error". Sume TTS 1.0 accepts 1 to 20,000 characters in one request, so a script ten times longer needs no splitting, as long as the audio stays under 1,200 seconds.

Deepgram's numbers are from its text to speech guide, read 2026-10-01; Sume's from the OpenAPI schema behind the API reference.

What are the two limits side by side?

Only the per-request text limit is compared; neither source gives a latency figure here.

Do I still need to chunk on Sume?

Only for very long text. At 20,000 characters or 1,200 seconds of speech, whichever comes first, split at sentence ends and send one request per chunk. The post on the 1,200-second limit shows how to size chapters.

What does a long single request look like?

An async job. The first response returns a status URL; poll it, or pass a webhook URL for a signed terminal callback.

curl -X POST https://api.sume.com/v1/tts-1.0/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: long-script-001" \
  -d '{
    "transcript": "Paste up to 20,000 characters here.",
    "avatar_handle": "@narrator",
    "mode": "async"
  }'

Why not just raise the limit on my side?

Deepgram's page gives no way to raise the 2,000-character cap, and the Sume schema's 20,000 is a fixed maximum too. Plan around the number you have: count characters before sending, since spaces and punctuation count toward Sume usage.