Python TTS cost calculator: MAI-Voice-2.1, Flash and Sume per job

This calculator prices a script on three options with the rates published on Microsoft's model page and in Sume's catalog. It reproduces Sume's cent rounding, so a 210-character script prints 1 cent and a 211-character script prints 2 cents.

The rates

Microsoft lists $22 per million characters for MAI-Voice-2.1 and $15 per million for Flash. Sume's TTS list rate is 38 micro-USD per character with a 1.25 multiplier, rounded up to whole cents per job, minimum 1 cent, maximum 95 cents at the 20,000-character request cap.

The code

Pure arithmetic: it needs no API key and makes no network call. Count characters as JavaScript would if you want to match the API; for plain ASCII scripts len() agrees.

import math

def sume_cents(n):
    return min(95, max(1, math.ceil(n * 38 * 1.25 / 10000 - 1e-9)))

def price(n):
    return {
        "sume_usd": sume_cents(n) / 100,
        "mai_usd": n * 22 / 1e6,
        "flash_usd": n * 15 / 1e6,
    }

for n in (60, 210, 211, 1000, 20000):
    print(n, price(n))

Expected output

• 60 characters: Sume $0.01, MAI $0.00132, Flash $0.0009
• 210 characters: Sume $0.01
• 211 characters: Sume $0.02
• 1,000 characters: Sume $0.05, MAI $0.022, Flash $0.015
• 20,000 characters: Sume $0.95, MAI $0.44, Flash $0.3

What it does not model

It ignores taxes, Microsoft's invoicing, any volume tier and anything beyond the TTS line. It also assumes the plain character count is what both vendors bill; Sume counts as JavaScript string length, so emoji and some accented text count differently.