Skip to main content

Metering

Gabber meters the following three types of usage:

  • Conversational Seconds
  • Voice Synthesis Seconds
  • Token Counts

Conversational Seconds​

Conversational seconds accrue when using any of Gabber's realtime voice-to-voice products.

Silence after 5 seconds is not metered and partial seconds are rounded up. Metering is tracked for both Human and AI only when the AI has finished their turn.

For example:

Human: Hey there. (0.4 seconds)
Silence: (2 seconds)
AI: How can I assist you today? (1.3 seconds)

Track Usage: Ceil(0.4 + Min(5,2) + 1.3) = 4 seconds

Silence (20 seconds)

Human: Goodbye AI, it was nice to meet you. (2.4 seconds)
Silence: (0.8 seconds)
AI: Goodbye (0.3 seconds)

Track Usage: Ceil(Min(5, 20) + 2.4 + Min(5, 0.8) + 0.3) = 9 seconds ***

NOTE: Token counts and voice synthesis seconds are not tracked when using the realtime voice-to-voice products. Any voice generation or LLM that Gabber uses is included in the conversational seconds metering and price.

Usage tokens can enforce limits on humans with the conversational_seconds limit type.

Voice Synthesis Seconds​

Voice synthesis seconds accrue when using the voice api to generate voice. They also accrue when using the /chat/completions endpoint if voice output is specified.

Token Count​

Tokens are metered when using the /chat/completions API with one of Gabber's provided LLMs. At this time, input and output tokens are metered the same.