One Sume TTS job: 20,000 characters is about 26 minutes of narration

A Sume text-to-speech request takes up to 20,000 characters. Using Unreal Speech's rule that one minute of audio is roughly 750 characters (read 2026-10-06), 20,000 characters is about 26.7 minutes of narration, priced at $0.95 at $0.0475 per 1,000 characters. So one job always covers the longest short-form video, with room to spare.

The platform ceilings come from Reap's guide (read 2026-10-06): YouTube Shorts up to 3 minutes, TikTok up to 10 minutes for most accounts and Instagram Reels up to 20 minutes, with Reels over 3 minutes not recommended to new audiences. Platforms change these limits, so check each before publishing.

Narration needed to fill each ceiling

When one job is not the right unit

A narration that fills the ceiling is rarely a good short. If you do narrate a long video, one job per chapter gives you audio you can retake without regenerating the whole track. Splitting a 15,000-character script into three 5,000-character jobs costs $0.72 instead of $0.72, because each job is rounded up separately.

Check the limit in code

The cap applies to the characters the job resolves to, including a newline where a script-source selection crosses a block. Go over and the API returns 422 tts_source_too_large (script source) or a validation error (literal transcript) before any charge.