MARS8 Pricing Calculator

Infrastructure Configuration

GPU Type Select GPU from the catalog

Workers Per GPU ?

Real-Time Factor (RTF) ?

On-Demand Price ($/hr per GPU)

Machine Type g2-standard-16

GPU Model NVIDIA L4

TTFB 580 ms

Hours Per Month ~730 average hours/month

Chars Per Second Per Stream ?

Utilization Target ?

Min GPUs (Autoscale Floor)

1-Year Commitment Discount 30% off on-demand

3-Year Commitment Discount 50% off on-demand

5-Year Commitment Discount 55% off on-demand

Define traffic distribution across peak, mid, and off-peak periods.

Phase	Fraction of Time	Concurrency (% of Peak)	Description
Off-peak			Nights / weekends
Mid			Typical daytime load
Peak			Busy hours near max
Total	1.00	Weighted Avg Load: 0.21

Peak Concurrency - Scenario A

Peak Concurrency - Scenario B

Monthly Volume 1 500M chars/mo

Monthly Volume 2 5.0B chars/mo

Monthly Volume 3 10.0B chars/mo

Scenario	GPUs (Static)	GPUs (Auto Avg)	Static $/mo (OD)	Static $/mo (1yr)	Static $/mo (3yr)	Static $/mo (5yr)	Auto $/mo (OD)	Auto $/mo (1yr)	Auto $/mo (3yr)	Auto $/mo (5yr)	$/1M Chars (Static)	$/1M Chars (Auto)

Annual Growth Rate 30% volume growth per year

Base Monthly Volume 500M chars/mo

Base Peak Concurrency

Static (On-Demand)

Autoscaled (On-Demand)

Year	Monthly Volume	Peak Concurrency	GPUs (Static)	GPUs (Auto Avg)	Static $/mo (OD)	Auto $/mo (OD)

MARS8 self-hosted cost vs. major TTS API providers (per 1M characters, autoscaled on-demand).

Scenario	MARS8 (Auto)	Google Std	Google Chirp HD	AWS Polly Neural	Azure Neural	OpenAI Std	OpenAI HD	ElevenLabs	Cartesia	Rime AI