How much will TTS actually cost?
Same job, different vendors, wildly different price tags. Estimate your monthly bill across closed-source, hosted OSS, and self-hosted options.
Your usage
That’s 1,800 minutes per month. About 30 hours.
Cheapest option
$0.900/mo
Self-hosted Kokoro (Modal/Runpod)
Most expensive
$540/mo
ElevenLabs Creator
OpenSpeech Cloud
OpenSpeech CloudComing soon
$27.00/mo
OSS models, hosted by us. Pay per minute, no commitments. · $0.015/min
Join the waitlist →
Closed-source
ElevenLabs Creator
$540/mo
Public Creator tier (~$0.30/min equivalent at typical speaking rate) · $0.300/min
ElevenLabs Pro
$324/mo
Pro / Scale tier (volume discount) · $0.180/min
OpenAI TTS-1
$162/mo
$15 per 1M chars ≈ $0.09/min at typical pace · $0.090/min
OSS on Replicate
Kokoro on Replicate
$5.40/mo
Tiny, fast, cheapest hosted OSS option · $0.0030/min
XTTS-v2 on Replicate
$27.00/mo
Voice cloning, popular default · $0.015/min
Chatterbox on Replicate
$45.00/mo
EL-tier quality, voice cloning · $0.025/min
Bark on Replicate
$144/mo
Slow + expressive; cold starts hurt · $0.080/min
Self-hosted OSS
Self-hosted Kokoro (Modal/Runpod)
$0.900/mo
Assumes ~60 generations/hr on A10G · $0.0005/min
Self-hosted XTTS / F5 / Chatterbox
$9.00/mo
Assumes moderate utilization on A10G/L4 · $0.0050/min
Self-hosted Bark / VibeVoice
$90.00/mo
Heavier models, A100 territory · $0.050/min
A few honest caveats
- Per-minute estimates assume average English speech (~150 wpm, ~900 characters/minute).
- Self-hosted costs assume reasonable GPU utilization. If your fleet sits idle 90% of the time, costs go up proportionally.
- Vendor prices change. Reviewed mid-2026 — for production planning, always confirm with the vendor.
- Latency, voice library size, voice cloning quality, and SLA are NOT priced in. ElevenLabs charges more partly because it does things the OSS models don’t yet match (5,000+ voices, sub-200ms TTFB, enterprise contracts).