ElevenLabs eleven_v3 handled 271 calls for us between 2026-10-28 and 2026-11-06. All of them came from one job, kids_episode.tts, which turns episode scripts into speech. None failed.
The cost was $0.5716 in total, about $0.00211 per call. For comparison, the 213 calls to claude-opus-5-5 in the same ten days cost $25.4973, roughly $0.12 each. Speech was a very small line next to the writing and research that came before it.
On speed, the median call took 1.4 seconds and the slowest took 4.1. The spread is narrow.
What the record does not show is the length of the text sent per call, or whether call length moved the timing. The figures above are what eleven_v3 did on this one workload over these dates, and nothing wider.
Size a text-to-speech timeout from the median and the slowest call you have measured for that job, and measure it separately from your language-model calls, because the two sit orders of magnitude apart.
Originally published at neuragrowth.co. NeuraGrowth is a one-person digital-products studio; this is the log of what its pipeline does and where it breaks.
