12-session meditation course narration on Sume TTS: $3.48 total

Twelve meditation sessions of 6,000 characters each cost $3.48 to narrate on Sume TTS 1.0 (read 2026-10-06). Each session is $0.285 at list, and every job rounds up to the cent, so a session bills $0.29 and the course bills 12 times that. The whole script is 72,000 characters, which is $3.42 before rounding, so the rounding adds six cents across the course.

A 6,000-character session reads at roughly the length of a short guided practice, but the length of your audio depends on the voice and the pace, so treat the character count as the price and the minutes as something you listen for. Sume TTS 1.0 accepts up to 20,000 characters in one job, which is $0.95, so a session never needs splitting for size.

Why the split barely matters, and when it does

The table shows the price varies by 18 cents across four very different ways to cut the course. Large jobs round once, small jobs round often, but at this scale the rounding is pennies. What changes more is how you work. One job per session gives you one file per session, which maps to one lesson in your app. Three sessions in a job gives you one long file that you would have to cut afterwards, and a mistake on page two means paying again for all 18,000 characters.

Short jobs help most when you rewrite. If a script editor changes one paragraph, a paragraph-sized job costs $0.03 to redo and a session-sized job costs $0.29. If you expect to edit, pick the smaller unit. If the script is locked, pick the session.

• Price by characters, including pause marks and stage directions that you leave in the text.
• Do a one-session pilot at $0.29 before committing the other 11.
• Failed jobs are refunded and cost nothing, so retry freely on a failure.
• Keep the script in version control, so each audio file matches a specific text.

Estimate the course before you submit

The function below prices a list of session lengths in characters, with the cent rounding Sume applies to each job.

def job_cents(chars: int) -> int:
    return -(-chars * 475 // 100_000)


def course(sessions: list) -> float:
    return sum(job_cents(n) for n in sessions) / 100


sessions = [6000] * 12
print(course(sessions))
print(course([12000] * 6))
print(course([600] * 120))

Pacing and pilot checks

Before you buy all twelve, read the pilot session aloud against a stopwatch and listen to the generated file once. You are checking three things: that proper nouns and Sanskrit terms are pronounced the way you want, that your punctuation produces the pauses you expect, and that the length of the finished file fits the slot in your course. The pilot costs $0.29, and it saves you from twelve files with the same fault.

If a word is mispronounced, fix it in the text, for example by respelling it, and regenerate only the affected session. Keep a short list of respellings in the script header so every later session uses them. A course is a set of files that must sound like one voice, and the cheapest way to keep that is one voice setting, one respelling list and one script style across all twelve jobs.

Budget for the rest of the course

Narration is the cheap part. If each session also gets a still or a short looping clip, that cost leads the budget, and it is better priced on its own. A calm background loop at the shortest Wan 3.0 duration (2 seconds at 480p) lists at $0.125 a clip, and the render that joins voice and picture is $0.10 per output minute, rounded up. Price them separately in the same way, then add the voice line.

Read the live numbers on the API pricing page and the credits page before you commit.