Baidu Xiling

What is Baidu Xiling?

Baidu Xiling is a platform for creating digital avatars, digital employees, and marketing content, launched by Baidu Smart Cloud. It offers functions such as digital avatar videos, live streaming, real-time interaction, corporate knowledge Q&A, image customization, voice cloning, and developer components. It provides a creation workspace that ordinary users can use, as well as APIs, SDKs, and MCPs for enterprises to integrate it into their systems.

Xi Ling is not a single \"digital voice generation tool.\" The current products consist of a marketing content creation platform and an open platform for digital employees; the former focuses on content and live streaming operations, while the latter emphasizes the combination of visuals, videos, interactions, and live streaming elements.

Public portraits and timbres

The official product documentation states that there are over 800 public avatars and more than 150 public sound effects available for use in videos, live broadcasts, and chat applications. The components, resolution options, and licensing restrictions that apply to each avatar may vary.

The fact that it is available publicly does not mean that it can be used for commercial purposes indefinitely at no cost. Before making a purchase, it is necessary to check the duration of use, the media types allowed, the regions, brand restrictions, and any limitations related to sensitive industries.

Digital human from photos

Photo digital avatars can be created quickly from a single image of a person, yielding 2D digital representations that can be used in videos featuring such avatars or for interactive applications. The basic customization service costs 20 yuan per transaction, while bulk orders are priced linearly based on the number of transactions.

The range of movements and perspectives in a single photo is limited; the image should be a clear front-facing shot, and permission for use as a portrait must be obtained.

2D small-sample digital human

2D digital avatars with a small sample size are created by training them using videos of real people. For generating generic lip movements, 1 to 4 minutes of video with the mouth closed is required; the current customization fee is 1,000 yuan per session.

Creating custom lip movements usually requires 4 to 8 minutes of speaking; the current cost is 2,000 yuan per session.

The lighting, clothing, movements, and background all affect the final outcome. Training avatars can be used in video, interactive, and live streaming components, but separate services must be purchased for each of these components.

Generative digital humans

For the basic generative version, training can be carried out by uploading a portrait photo; the process is said to take around 40 minutes. The cost is currently 20 yuan per session, and it is mainly used for video components. The advanced version offers more natural movements and interaction patterns, and it supports both video and interactive functions; its cost is currently 2000 yuan per session.

The basic version is not suitable for real-time interactions that require a seamless transition between listening mode and playback mode; it is advisable to view sample videos before making a choice.

3D digital human

Xi Ling offers text-to-3D generation, 3D cartoon creation, and customization of corporate models. 3D digital avatars can be used in videos and interactive applications, and they support various screen formats as well as half-body and full-body representations.

The rapid version of 3D customization currently costs 60,000 yuan per portrait. If the company provides the original artwork or model, it still needs to be adapted, and this process can take several weeks.

2D hyper-realism and endpoint rendering

2D hyper-realistic digital avatars are trained using static and video content for use in cloud-based rendering interactions; 2D-rendered digital avatars, on the other hand, operate on end devices and are suitable for use on large screens, in apps, as well as in scenarios that require offline operation or low latency.

End rendering requires an SDK, a license, device compatibility, and TTS resources; it is not possible to go live simply by purchasing avatars alone.

123 Digital Human Video

123 Digital Human is one of Xiling’s video generation tools; it enables the creation of digital avatars that can speak, using visuals, voices, and text, and it offers official APIs. It is suitable for generating large quantities of news content, educational materials, and marketing content.

Batch tasks require handling of asynchronous statuses, callbacks, retry on failure, and result downloading; the output scripts also need to undergo fact-checking and compliance audits.

Video creation workspace

The marketing content platform currently lists a variety of creation methods, including turning product images into videos, video translation, converting inspiration into videos, creating curated videos, AI-driven advertising, PPT generation, AI business cards, and viral videos.

Workbenches lower the barriers to video editing, but AI-generated products, subtitles, prices, and character movements may contain errors; manual verification is still required before official release.

Product images and AI-driven advertising

Product image-to-video tools enable the conversion of static product materials into marketing videos, while AI advertising facilitates the rapid creation of content based on key features, scripts, and digital avatars; these tools are suitable for use in e-commerce and retail operations.

The model may alter the text on the packaging and the details of the products. The effectiveness of the ads, prices, inventory levels, and expiration dates of any offers must be based on the actual business conditions.

Video translation

Xi Ling offers a video translation interface that can handle audio, subtitles, dubbed versions in the target language, as well as the orientation of character lips; it is suitable for cross-border marketing and training purposes.

Technical terms, cultural expressions, and disclaimers require review by a native speaker, and permission must be obtained to use cloned voices.

Digital human live streaming

The live streaming component enables 2D digital avatars with small sample datasets to deliver continuous broadcasts, and it integrates smart script generation, interactive messages from viewers, as well as the customization of the live streaming environment. The marketing platform also allows for authorization of platforms such as Meituan, Taobao, and JD.com, along with the ability to view live streaming data.

Even in the case of unattended live streaming, it is necessary to comply with the platform’s rules and the requirements regarding AI-generated content labels. Human intervention is required for handling products and responding to interactions.

Real-time dialogue digital human

Interactive components combine speech recognition, large models, knowledge bases, speech synthesis, and digital avatar rendering to enable real-time conversations. Users can create digital avatars with specific knowledge, skills, and personalities.

Real-time applications have to deal with latency, interruptions, sensitive issues, hallucinations, and network abnormalities. High-risk scenarios such as those in healthcare and finance require human intervention.

Knowledge base and agents

The platform allows users to upload documents in order to create custom knowledge bases, enabling digital avatars to answer questions by drawing on general knowledge as well as industry- and company-specific information. It is suitable for use in corporate exhibition halls, customer service, training, and guided tours.

The knowledge base requires version updates, as well as verification of permissions and references. Answers provided by large models cannot replace official policies and contracts.

Voice cloning

The LITE version of voice cloning costs 10 yuan per use, while the standard version costs 200 yuan per use; both options come with service packages of 10 or 100 uses. The TTS system that is generated as a result of training can be used in videos featuring digital avatars, as well as in interactive and live streaming applications.

Explicit consent from the speaker must be obtained before cloning, and the Xi Ling Cloning Agreement must be followed. It is not allowed to impersonate or forge authorization.

Speech synthesis and SSML

The platform offers a text-to-speech API as well as SSML controls, which enable the use of text to drive digital avatars and allow for adjustments to pauses, speech speed, and pronunciation. End-rendered interactions typically require character packets and concurrency.

Currently, the TTS character package costs 100 yuan for 100,000 characters and 1,000 yuan for 1 million characters, with a validity period of one year; concurrent capacity expansion costs 40 yuan per month or 400 yuan per year.

Video generation duration package

The video components of the open platform are billed based on the duration of production, with a validity period of usually one year. The cost for a digital avatar based on a photo is 45 yuan for 10 minutes; for 2D versions and 123 digital avatars, it is 30 yuan for 10 minutes. A high-quality 2D avatar costs 40 yuan, while a 3D digital avatar costs 50 yuan.

100 minutes and 1000 minutes are billed at the same rate. The shared resource pool does not offer dedicated concurrency, and the speed during peak times is affected by the shared resources.

Video-specific concurrency

The price for digital avatars based on photos is 4,500 yuan per month or 45,000 yuan per year; for 2D versions with small sample sizes, high-quality 2D versions, 3D versions, and 123 digital avatars, the cost is usually 2,000 yuan per month or 20,000 yuan per year.

For video billing, you can opt for just a duration package, or a combination of a duration package and dedicated concurrency. For small-scale tasks, the public pool can be used first.

Price and version comparison

Interactive cloud rendering pricing

For 2D cloud rendering, the cost is 800 yuan per 7 days, 2,400 yuan per month, and 24,000 yuan per year; for 3D rendering, the costs are 1,200 yuan per 7 days, 3,600 yuan per month, and 36,000 yuan per year.

Interactive applications also require resources such as graphics, voice, large models, and the internet; therefore, the total cost cannot be determined solely based on the price per concurrent connection.

Price of the client-side rendering SDK

The license for rendering photo-based digital avatars on the H5 platform currently costs 10,000 yuan per month or 100,000 yuan per year. For 2D rendering, the device license costs 3,600 yuan per unit, while the application license costs 36,000 yuan per unit.

Application authorization applies only to a limited number of devices on either Android or iOS systems.

3D rendering is billed on a per-concurrency basis; the current rates are 120 yuan for 7 days, 360 yuan per month, and 3600 yuan per year. The specific compatibility requirements and device limits are specified in the contract.

Price of live streaming components

The API for live streaming of 2D digital avatars with a small sample size offers a concurrent usage rate of 700 yuan per 7 days, 2,000 yuan per month, or 20,000 yuan per year; in addition, access to the standard portrait library is available for a limited time.

Live streaming also requires image training, content, voice quality, and streaming resources; purchasing these elements separately does not constitute a complete solution.

Why buy in bundles?

The official price guide specifies that purchasing individual products such as avatars, voiceovers, or interactive elements on their own generally does not result in a complete final application. Avatars determine \"who appears\", duration or concurrency parameters determine \"how it is generated\", while TTS and large models handle \"what to say and how to respond\".

Before making a purchase, it is necessary to draw up the complete workflow and select components based on video, live streaming, or interactive scenarios.

APIs and SDKs

Xi Ling Document offers file uploading, custom avatar creation, voice cloning, video synthesis, image-to-video conversion, video translation, video editing, cloud rendering, session management, streaming, and SDKs for multiple platforms.

Production access requires protection for authentication, verification callbacks, limitation of concurrent operations, saving of task results, and handling of License updates.

Digital Human MCP

The open platform is compatible with MCP, allowing large-model applications that support MCP to upload materials, create digital avatars, synthesize voices, and perform searches, while also enabling the use of Xi Ling’s capabilities within conversations.

Using MCP still requires a platform account, relevant permissions, and a cloning protocol; it does not mean that it can be used free of charge or locally.

Is it open source?

The Xi Ling platform, digital human models, rendering services, and SDKs are closed-source commercial products; no complete core source code or model weights that have been made publicly available by the developers exist.

The fact that GitHub supports third-party digital human projects that integrate Baidu Xiling does not mean that Xiling itself is open-source.

Baidu Xiling Usage Guide

Complete a basic task.

1. Identify the audience, platform, format, duration, and the information that needs to be conveyed;
2. Prepare scripts, shots, or reference materials that can be used in Baidu Xiling;
3. Select a public portrait and voice tone to create a low-cost preview;
4. Use photo-based digital avatars to adjust the visuals, rhythm, subtitles, and audio;
5. Check each frame for characters, text, logos, lip movements, and factual accuracy;
6. Export in the desired format after confirming the licensing for music, portraits, and materials;

Create reusable professional workflows

1. Create scripts, shot lists, brand assets, and a list of elements that are prohibited from use;
2. Uniform parameters are saved for public portraits and voice tones, as well as for digital humans based on photos and 2D low-resolution digital humans;
3. First, use representative shots to test the model and the quota;
4. Transfer the failed shots to manual editing or regenerate them;
5. Uniformize subtitles, volume, colors, and end credits;
6. Record the version and reviewer before publishing in batches;

Which users is it suitable for?

• A team responsible for creating bulk digital voiceovers and courses;
• Companies that carry out e-commerce live broadcasts and marketing videos;
• Organizations that create exhibition halls, provide customer service, offer guidance, and train digital employees;
• Developers who need 2D, 3D, or sign language digital human SDKs;
• Platforms that require voice cloning, API, and MCP integration.

Product advantages

• Video, live streaming, interactive, and agent scenarios are all fully supported;
• There are a wealth of public portrait and timbre resources;
• There are various forms such as photos, 2D, 3D, and end-rendered versions;
• The official price list is broken down by duration, concurrency, and license;
• Provides APIs, SDKs, and MCP;
• It can be combined with Baidu Cloud’s capabilities and large-model technologies.

Restrictions and Precautions

• There are many product components, and the procurement combination is complex;
• The image, generation time, concurrency, interactive licenses, TTS, large models, and cloud resources may be charged separately;
• Digital avatars, voices, and marketing content are related to issues such as portraits, voices, copyright, and advertising compliance.
• All AI-generated outputs require manual review;

Frequently Asked Questions

Is Baidu Xiling free?

By logging in, you can obtain certain trial benefits; once the free quota is used up, it is necessary to purchase additional components from the component store. For official enterprise use, charging is based on the number of components used.

How much does a digital human based on a photo cost?

The customization of photo images currently costs 20 yuan per service, while the creation of a video lasting 10 minutes costs 45 yuan; for interactive applications, components such as SDK and TTS are required.

How much does voice cloning cost?

The LITE version costs 10 yuan per use, while the standard version costs 200 yuan per use; both versions offer bulk service packages.

Is it possible to conduct live broadcasts with digital avatars?

Yes, the 2D small-sample live streaming API currently comes at a cost of around 2000 yuan per month; in addition, resources such as avatars and content are required.

Does Baidu Xiling have an API?

It offers complete APIs, H5, and mobile SDKs, as well as support for MCP, covering video, interaction, live streaming, and customization.

Is Baidu Xiling open source?

It is not open source; the complete source code for the core platform, models, and rendering services is not made available.