One 600-character voiceover in 12 languages: TTS cost by vendor

A common production job: one product voiceover, translated into 12 languages. Assume each version is about 600 characters. This is an assumption; translations run longer or shorter than the source, and scripts in some languages take fewer characters for the same meaning. Re-count after translating. Twelve versions of 600 characters is 7,200 characters.

TTS is billed per character at each vendor's rate. OpenAI lists tts-1 at $15 and tts-1-hd at $30 per 1M characters. ElevenLabs lists $0.04 per 1,000 for Flash/Turbo, $0.08 for v3, and a v4 launch price of $0.022 through October 12 (list $0.08), all read 2026-10-03. Sume is $0.0475 per 1,000 characters.

The bill

Each row is the per-1,000 price times 7.2.

Scale it to a catalog

For 250 products localized to 12 languages, that is 3,000 voice lines, or 1.8 million characters at 600 each. On Sume that is $85.50. On OpenAI tts-1 it is $27.00 and on ElevenLabs v3 it is $144.00. In every case the voice is a small fraction of the cost of the video it will sit under: a 15-second clip on Sume at $0.10 a second is $1.50 on its own.

The multiplier that matters here is retakes. A wrong pronunciation in one language means paying for that line again, so budget a margin, say 10 percent, and check the audio for the languages you cannot read before the batch ships.

Compute your own

This snippet prices any set of lines at the Sume rate.

lines = 12
chars_per_line = 600
rate_per_1k = 0.0475  # Sume TTS
total_chars = lines * chars_per_line
print(f'{total_chars} characters -> ${total_chars / 1000 * rate_per_1k:.4f}')