GradiumBENCHMARKS

Read on November 19, 2026

Frequently asked questions

How many credits do I need?
TTS: one minute of audio typically requires 750 characters (750 credits). STT: 3 credits per second. Speech-to-Speech Translation: 30 credits per second. Speech-to-Text Translation: 4 credits per second.

Source : gradium.ai/pricing-faq

What languages do you support?
English, French, Spanish, Portuguese, and German, with more languages in development.

Source : gradium.ai/pricing-faq

What is the maximum session duration?
Pricing FAQ: up to 300 seconds. Docs FAQ and Limits: up to 3000 seconds (7200 on L); free tier limited to 1,500 characters per session. TTS page: five minutes.

Source : gradium.ai/pricing-faq · docs.gradium.ai/guides/faq.md · docs.gradium.ai/guides/limits.md · gradium.ai/text-to-speech

What happens when I subscribe to a plan?
The plan activates immediately and the account is credited with the monthly quota.

Source : gradium.ai/pricing-faq

What is the billing interval?
Monthly or yearly, starting from the day you subscribe; cancel any time.

Source : gradium.ai/pricing-faq

What happens when I upgrade during the month?
New plan activates immediately with its credits; remaining credits from the current plan are retained.

Source : gradium.ai/pricing-faq

What happens when I downgrade during the month?
Current plan ends at the close of the billing cycle; new plan starts the next cycle.

Source : gradium.ai/pricing-faq

What happens when I have used all my credits?
Most plans offer usage-based billing (pay as you go) to be activated in settings; otherwise model access stops.

Source : gradium.ai/pricing-faq

What happens to unused credits at the end of the month?
Plan credits cannot roll over; remaining pay-as-you-go credits roll over.

Source : gradium.ai/pricing-faq

What is an Instant Voice Clone?
A replica from typically 10 seconds of reference audio; up to 1,000 voices depending on plan; voice owner's consent required.

Source : gradium.ai/pricing-faq

What is a Pro Voice Clone?
A high-fidelity voice model fine-tuned on a large dataset of your audio (minimum 30 minutes, 2 hours recommended); Plan M and above.

Source : gradium.ai/pricing-faq

How do you charge for Pro Voice Clones?
One-time training fee of 1M credits plus 1.2 credits per character.

Source : gradium.ai/pricing-faq

Where is my data stored?
In the region closest to you: US requests processed and stored in the US, EU requests in the EU.

Source : gradium.ai/pricing-faq

How do you measure latency?
From the moment a caller hears sound; median 50 ms with no leading silence; P25-P75 gap 5 ms.

Source : gradium.ai/text-to-speech

Do languages cost extra?
No. All five languages are included on every plan at no extra charge.

Source : gradium.ai/text-to-speech

Can we run it inside our own infrastructure?
Our cloud by default, AWS Marketplace, or your own VPC as a SageMaker image; dedicated, self-hosted and on-premise on enterprise plans.

Source : gradium.ai/text-to-speech · gradium.ai/api

Can a clone made on the free plan be used commercially?
No. The free plan is for internal, non-commercial use; commercial use starts on XS.

Source : gradium.ai/voice-cloning

Do you use my data to train your models?
No, except content you authorise or choose to share with Gradium.

Source : gradium.ai/legal-faq

Can I use Phonon commercially?
Phonon 31m is Apache 2.0; Phonon 40m models are CC BY-NC 4.0 and need a separate commercial license.

Source : gradium.ai/phonon

How fast is live translation?
About 3 seconds on average end to end, across all language pairs.

Source : gradium.ai/live-translation