Billing items, units, and methods for Intelligent Speech Interaction-Intelligent Speech Interaction(ISI)-阿里云帮助中心

Intelligent Speech Interaction provides a free trial and Commercial Edition. Speech services are billed by audio duration, number of calls, or number of synthesized characters.

Free trial and Commercial Edition

Intelligent Speech Interaction provides a free trial and Commercial Edition. The free trial supports limited use of speech services. If the free trial does not meet your business requirements, activate Commercial Edition or purchase a resource plan.

Category	Free trial	Commercial Edition
Billing rules	New users receive a three-month free trial after activating the service. Warning If you upgrade to Commercial Edition during the three-month free trial and then downgrade to the free trial, the available concurrency is reset to 0 and the service becomes unavailable. Proceed with caution. If the system detects repeated free trial activations in violation of the rules and malicious requests to speech services, Alibaba Cloud may take action without prior notice. You are responsible for the resulting consequences.	After you activate Commercial Edition, Intelligent Speech Interaction supports pay-as-you-go (postpaid) and resource plans (prepaid). For pricing details and upgrade instructions, see Billing.
Concurrency and QPS limits	Only some services in the free trial support low concurrency. To use higher concurrency, upgrade to Commercial Edition. If the concurrency of Commercial Edition still does not meet your business requirements, increase the concurrency quota. For concurrency details and upgrade instructions, see Concurrency and QPS limits.
Supported services	Speech recognition: Recording File Recognition, Real-time Speech Recognition, Short Sentence Recognition Speech synthesis: Speech Synthesis Speech analytics: Speaker Recognition, Sound Event Detection, Gender Recognition, Language Identification	Speech recognition: Recording File Recognition, Real-time Speech Recognition, Short Sentence Recognition, Recording File Recognition (Express Edition), Recording File Recognition (Off-Peak Edition) Speech synthesis: Speech Synthesis, Long Text-to-Speech, Streaming Text-to-Speech (CosyVoice large language model), VoiceChat Speech analytics: Speaker Recognition, Sound Event Detection, Gender Recognition, Language Identification

Billing items

The Commercial Edition of Intelligent Speech Interaction is billed based on the volume of speech data processed.

Additional charges apply when you use add-ons such as excess concurrent lines.

Speech data processing

The billing rules are as follows:

Duration-based billing: The audio duration of each call is accumulated and rounded down to the nearest second. For example, 22.8 seconds is billed as 22 seconds.
Per-call billing: Failed calls are not counted.
Multi-channel billing for recording files:
- 8 kHz dual-channel audio is billed as single-channel audio based on audio duration.
- 16 kHz multi-channel audio is billed based on number of channels × audio duration.
Real-time Speech Recognition is billed based on the streaming duration of each channel. If the two channels of a call are streamed separately, their durations are added together.
The number of billable Speech Synthesis calls is calculated based on the number of UTF-8 encoded characters in each request. A request of up to 100 characters counts as one billable call. Each additional 100 characters adds one billable call. A single request can contain up to 300 characters.

One Chinese character counts as two effective characters. Each English letter and full-width or half-width punctuation mark counts as one effective character. SSML tags are excluded from the character count.

For example, 102 characters count as two billable calls, and 201 characters count as three billable calls.

Category	Service	Billing unit	Billing method
Speech recognition	Real-time Speech Recognition	Billed by audio duration	Pay-as-you-go (postpaid) or resource plan (prepaid)
	Short Sentence Recognition	Billed by number of calls
	Recording File Recognition	Billed by recording duration
	Recording File Recognition (Express Edition)	Billed by recording duration
	Recording File Recognition (Off-Peak Edition)	Billed by recording duration
	Alibaba Cloud Model Studio voice model service	Billed by audio duration	Pay-as-you-go (postpaid)
Speech synthesis	Speech Synthesis	Billed by number of calls	Pay-as-you-go (postpaid) or resource plan (prepaid)
Speech synthesis	Long Text-to-Speech	Billed by number of synthesized characters
Speech analytics	Sound Event Detection	Billed by recording duration
	Speaker Recognition	Billed by number of calls
	Gender Recognition	Billed by number of calls
	Language Identification	Billed by number of calls

Additional product fees

Service

Description

Excess concurrent lines

Commercial Edition provides 200 default concurrent calls for Short Sentence Recognition and Real-time Speech Recognition and 10 for Recording File Recognition (Express Edition). If your business requires more, you can purchase additional concurrent lines.

For pricing details, see Billing.

For purchase instructions, see Concurrency and QPS limits.

Custom language model tool

The custom language model tool helps improve recognition accuracy in specific domains.

Custom language models are now available free of charge. Each user can create up to 10 models.