Billing and Pricing

The AI Real-Time Conversation Engine provides real-time multimodal AI conversation capabilities for smart devices. Usage is billed by effective call duration, with minutes as the base billing unit.

1. Billing Method

The AI Real-Time Conversation Engine is billed by effective call duration, with minutes as the base billing unit.

Audio and video calls are billed separately according to the actual call mode and are not charged twice for the same period. Advanced models are priced on demand.

2. Standard Billing Items

Standard billing item	Standard price in Chinese mainland	Billing condition
Audio call service	CNY 0.13/minute	A device has established a conversational connection with an AI agent and entered the running state, without providing actual video input
Video call service	CNY 0.40/minute	The device actually provides video input to the AI and visual understanding is enabled
Advanced model service	On-demand pricing	An advanced model or dedicated model solution is actually used

There is currently no single publicly listed price for overseas and other service regions. Refer to your contract or order for the applicable price.

The AI Real-Time Conversation Engine depends on the TiRTC audio and video channel service. You must activate TiRTC before use, and TiRTC fees are billed separately.

3. Billing Item Details

3.1 Audio Call Service

The audio call service is billed for the effective period during which a device has established a conversational connection with an AI agent, entered the running state, and is not providing actual video input. This billing item covers real-time voice interaction using standard models.

3.2 Video Call Service

The video call service is billed only for the effective period during which video is actually provided to the AI and visual understanding is enabled. Its price includes the audio call capability for the same period, so the audio call service is not charged again. If video capability is enabled but no valid video input is produced, the period is still billed as an audio call.

3.3 Advanced Model Service

The advanced model service incurs charges only when an advanced model or dedicated model solution is actually used. Available models, billing methods, service scope, and prices are subject to the applicable contract or order.

Special models or dedicated solutions outside the standard support scope are quoted in the contract or order.

3.4 Capabilities Without Separate Charges and Other Charges

• Knowledge bases, long-term memory, the basic MCP toolchain, and standard voice cloning are not separate billing items. Their available scope and limits are subject to the product description.
• The provided public map service is currently not charged separately. If its billing rules change, refer to the updated billing description, console notification, or contract.

4. Usage Calculation Rules

4.1 Start and End Times

Effective duration starts when the device establishes a conversational connection with an AI agent and enters the running state. It ends when the session is actively ended, the connection times out, or the agent stops running.

Creating or saving an agent configuration, leaving a device disconnected, or maintaining only a network heartbeat does not incur AI Real-Time Conversation Engine fees.

4.2 Aggregation Rules

1. For each effective call, the actual active duration of every standard billing item is recorded in seconds.
2. The effective seconds for each billing item are aggregated separately by customer account, application, service region, device, and individual call.
3. Within an individual call, the aggregated seconds for each billing item are divided by 60 and rounded up separately to whole minutes.
4. If the same device makes multiple calls, each call is metered and rounded separately.

For example, if the same device makes 10 standard audio calls of 6 seconds each, every call is billed as 1 minute, for a total of 10 audio call minutes.

4.3 Configuration Changes and Exceptional Duration

• The video call service is billed only for periods when video input is actually provided and visual understanding is enabled.
• The advanced model service is billed only for periods when an advanced model is actually used.
• If video or model configuration changes during a session, the system updates the active billing items at the change time and creates billing segments.
• By default, billing for a session ends after 60 consecutive seconds with no speech, no effective tool execution, and no AI playback. The actual timeout is subject to the product configuration.
• Periods confirmed unavailable due to a Tange platform failure are not billed.

5. Fee Calculation and Bills

5.1 Fee Formula

AI Real-Time Conversation Engine fee = audio call minutes × CNY 0.13 + video call minutes × CNY 0.40 + advanced model service fees specified in the contract or order

Total customer fee = AI Real-Time Conversation Engine fee + TiRTC audio and video channel service fee + other separately billed service fees

Audio and video calls are billed in segments according to the actual call mode, without duplicate charges for the same period. Advanced model services are settled according to the contract or order.

5.2 Usage Details

Bills and usage details are generated from the actual settlement results and include billing items, effective duration, settlement unit price, and fees.

Knowledge bases, memory, basic MCP, standard voice cloning, and public map capabilities do not form separate billing items.

6. Account and Settlement

6.1 Top-Up and Deduction

• The AI Real-Time Conversation Engine uses a prepaid enterprise account model: top up first, then consume.
• After a customer tops up, the amount is credited to the enterprise account balance.
• Each standard billing item is deducted directly from the enterprise account balance according to actual usage.
• Pending charges caused by the delay between metering and bill generation are deducted first after the customer tops up.

6.2 Bills and Overdue Balance Notifications

• The system records actual usage and fees. Customers can view billing items, settled usage, and fees on their bills.
• When the enterprise account balance is insufficient to cover the amount due, the account becomes overdue and the system sends overdue balance alerts.

6.3 Overdue Balance Handling

An overdue account enters a grace period followed by a retention period:

1. Grace period (1 day): The grace period lasts for 24 consecutive hours from the time the account becomes overdue. During this period, the AI Real-Time Conversation Engine remains available, actual usage continues to be metered and settled, and the system continues to send overdue balance alerts.
2. Service suspension and retention period (15 days): If the overdue balance remains unpaid when the grace period ends, the system stops establishing new AI conversation sessions and ends AI conversation sessions that are still running. Agent configurations, knowledge bases, memory, and other related business data are frozen and retained for 15 calendar days. No new AI Real-Time Conversation Engine usage fees are incurred during the retention period.
3. Recovery during the retention period: After the customer tops up, the system first deducts pending settlement charges. Once the overdue balance is fully paid, the AI Real-Time Conversation Engine is automatically restored, and configurations and data that have not been deleted during the retention period become available again.
4. End of the retention period: If the overdue balance remains unpaid at the end of the 15-day retention period, related agent configurations, knowledge bases, memory, and other business data are deleted and cannot be recovered.

The TiRTC audio and video channel service and other related products are billed independently. Their overdue balance handling and data retention rules are subject to their respective billing descriptions. Suspending the AI Real-Time Conversation Engine does not automatically stop other activated services.

7. Additional Notes

• Prices listed here are standard prices for Chinese mainland. Custom development, dedicated models, private deployment, and services outside the standard support scope are quoted separately.
• If product prices or billing rules change, refer to the applicable billing description, console notification, or contract.