ElevenLabs character limits by model vs Sume TTS 20,000

ElevenLabs sets a different character limit per model: Eleven v4 10,000, Eleven v3 5,000 and Flash v2.5 40,000, per its models page. Sume TTS 1.0 has one text cap, 20,000 characters in transcript, plus a separate cap on the synthesized audio length.

Vendor numbers are from the ElevenLabs models page read 2026-10-01; v4 and v4 Turbo shipped 2026-09-28. Sume numbers are from the OpenAPI schema and Timeline audio.

What are the ElevenLabs limits?

The models page lists the limit on each model card. The numbers below are as written there.

What is the Sume text limit?

The transcript field in the TTS 1.0 request is 1 to 20,000 characters, and the schema says spaces and punctuation count toward usage. There is no per-engine figure because TTS 1.0 has no engine picker.

A separate rule applies to audio: synthesized audio longer than 1200 seconds fails with tts_duration_exceeded. So a script can be under the character cap and still be too long in speech. Check both.

How do I handle a script longer than the cap?

Split the script into chapters, synthesize each, and join them. Timeline audio joins Sume-hosted audio into one gapless file and parts[] takes 1 to 20 ordered parts. The chapter approach is covered in TTS 1200-second limit for audiobook chapters.

Which number should I check first?

Compare your longest script to the model you would use on ElevenLabs, then to 20,000 characters and 1200 audio seconds on Sume. Cut at paragraph boundaries so each part ends on a natural pause.