Starry Future – SD Model Collection

What is the Star Han Future SD model set like now?

“StarFuture–SD Model Set” was the name of the directory used in the early stages of this platform for providing cloud deployment, LoRA, and API versions based on Stable Diffusion. The product has now evolved into a comprehensive AI application market operated by Beijing StarFuture Network Technology Co., Ltd.; its public interface offers services related to voice, AI agents, writing, data processing, and multi-modal applications.

The old Stable Diffusion section now redirects to the exploration page, and the public application list does not show any SD versions that can be deployed immediately. Users should consider this entry as an explanation of the current status of the old entry point; they should not assume that the models are still available based on their former names.

Current operational status and product changes

This change does not mean that the company has stopped operating; rather, its product range and access methods have been adjusted. If Stable Diffusion is specifically needed, one should first search the market to find out details about the application, information on the deployment button, the version of the model, support for LoRA, and the current price.

Current main functions

Cloud AI Application Market

• The app store displays the app’s name, description, deployment method, and the amount of points consumed every 10 minutes; users can start using these apps without having to set up servers themselves.
• The list published this time includes applications for text-to-speech generation, background removal from images and videos, artificial agents, knowledge bases, essay grading, academic assistance, chatting, writing, as well as chart and data processing tools.
• Applications can originate from open-source projects, third-party models, or integrated platform versions; the login credentials and data rules may vary depending on the application.
• The indication of “cloud deployment” on the card simply means that the platform is responsible for hosting the application; it does not imply that the application code, model weights, or service data are all owned by Xinghan Future.

Application installation, suspension, and restoration

• After selecting an application, you can go to the details page to view its resource usage; once confirmed, it can be installed or deployed on your personal application desktop.
• Charges are applied based on the duration of use; users can set an automatic pause time to prevent continuous consumption due to forgetting to turn it off.
• Some applications have their own login systems; users may need to use the initial accounts provided by the application or create their own accounts within it.
• Pause, resume, and uninstall are distinct operations; after a task is completed, it is necessary to ensure that billing for the instance has stopped and to check the transaction history.

Pay-per-use package

• The pay-per-use package emphasizes the elimination of the need for installation, and it offers bundled subscriptions for the number of times a specific application’s large models can be used.
• The academic package includes six applications: ChatExcel, Novel, AIAcademic, AITutor, AIBot, and DeepSeek online assistant.
• The multimodal package includes two applications; the specific list can be viewed by switching to the package on the purchase page.
• The number of package uses and the points used in the app market are not the same billing units; it is necessary to determine before making a purchase whether the charges are based on the number of uses or on the amount of time used.

Multi-model API

• The platform provides interfaces for chat completion, text vectors, and model listing, which developers can use after generating an API Key in their account.
• The chat interface supports parameters such as system, user, assistant roles, multi-turn messaging, streaming responses, maximum output length, temperature, and top_p.
• The vector interface converts Chinese and English text into vectors, which is suitable for semantic search, retrieval enhancement, and similarity calculation.
• The model list interface is used to query the models that are available for use at the moment; developers should not rely solely on the static model names listed in the documentation.

Current application example

The list of applications, their deployment locations, and energy consumption levels will change in accordance with the resource costs; the table above shows only the information available on the public page at the time of this verification. Before actual use, it is necessary to reopen the details page to check whether the application is online, what the billing method is, whether automatic suspension is enabled, and what the data processing rules are.

Usage tutorial

1. Register or log in to the Xinghan Future AI Application Platform, then go to the Explore page to view the available applications at the moment.
2. Select Voice, Agent, Multimodal, Writing, or Data Applications based on the task; do not rely solely on the old names of the “SD model sets” to make your choice.
3. Open the application details to check the deployment method, power consumption per 10 minutes, input restrictions, initial account, and model version.
4. Based on the points recharged for short-term testing budgets, or compare the list of applications, the number of uses, and the valid conditions within the pay-per-use packages.
5. Install or launch the application, set a shorter automatic pause time, and then submit small-scale inputs to verify the quality of the output.
6. Once the task is completed, pause and uninstall the instances that are no longer in use, and check to ensure that the number of checkpoints and the application workflow stop increasing.
7. When system integration is required, go to API Key Management to create a key and first check the list of models to identify the current model.
8. Before connecting to production, implement key isolation, rate limiting, retry mechanisms, balance alerts, and output auditing.

Points, packages, and API prices

There are three different billing mechanisms: usage based on duration, number of package uses, and per-API call; these cannot be converted into one another to represent the same level of usage. Before making a purchase, it is necessary to check the validity period of the package, the applications for which it is applicable, whether failed calls incur charges, the validity period of the credits, and the refund policies.

The current public page does not provide a unified refund policy, nor does it outline the conditions for refunds in cases of unused points, unused package visits, or application errors. It is recommended to make small deposits first and keep records of the orders, points usage, and any abnormal occurrences.

API integration process and restrictions

1. Log in to the platform and go to API Key management to create a key intended for use only with the current project.
2. First, query the list of models to obtain the names of the chat or vector models that are available for use at the moment.
3. Messages, models, streaming switches, and generation parameters are organized through the chat completion interface, and authentication is carried out using Bearer tokens.
4. When a search in the knowledge base is required, the vector interface is invoked to convert the array of texts into embedding vectors.
5. Record the model, parameters, usage amount, response code, and cost for each request, as well as the codes and messages contained in any error objects.
6. The server stores the keys securely; in the event of a leak, they are immediately replaced, and the full keys are not saved in the frontend, public repositories, or logs.

• The model context, output limitations, and unit prices listed in the document may vary depending on the supplier; the list of models only shows their names and cannot replace actual testing.
• The temperature and top_p values vary across different models; it is therefore not possible to use the same parameter settings for all providers.
• The interface uses a data structure similar to that of OpenAI’s chat completion feature, but this does not imply full compatibility with all functions of the OpenAI SDK.

Teams and use cases

• Individual creators: They can use voice, image, video, writing, and charting tools for short periods of time, without the need to maintain a GPU environment on a long-term basis.
• Developers: Quickly test product prototypes using multi-model APIs or platform-hosted applications.
• Education and research users: Use the academic package for assisting with papers, analyzing tables, answering questions to supervisors, and organizing materials.
• Enterprise Team: After a team is created, applications, members, and point histories can be managed in a unified manner; the person who created the team can install and uninstall applications.
• Regular team members can restore and pause applications as well as view metrics, but they cannot install new applications; the permissions should be aligned with the internal roles.

Open-source status and license

• The presence of \"open-source applications\" in the market does not mean that the StarHan Future AI application platform as a whole is open source, nor does it imply that all model weights can be used for commercial purposes.
• The GitHub organization StarHan Future has made CostPilot and InfraPilot available; both repository pages indicate that they are licensed under the Apache-2.0 license.
• These two repositories are intended for cloud cost management and the Serverless platform respectively; they cannot be used as proof of an SD model, an application marketplace frontend, or open-source third-party applications.
• Each application should separately verify the licenses for its code repository, model weights, data, plugins, and container images.
• When calling third-party model APIs, it is also necessary to comply with the terms set by the model provider, as well as its content policies and commercial restrictions.

Notes on privacy, security, and copyright

• The public page of the current application platform does not contain a complete user agreement or privacy policy; details regarding data storage, deletion, sharing, and cross-border processing are not yet available.
• Different cloud applications may have their own accounts, databases, and logs; the privacy rules related to platform accounts cannot automatically apply to the data contained within those applications.
• Do not upload identification documents, customer information, unpublished code, trade secrets, or audio and images for which you do not have permission to use, unless the relevant deployment and data rules have been confirmed.
• API Keys, the initial password for an application, and team invitation details are all sensitive credentials that should be used with the minimum level of access required, and their values should be updated regularly.
• Creating images, audio, video, and text may involve training data, rights related to individuals, trademarks, and materials from third parties; separate review is required before using them for commercial purposes.
• Point consumption and automatic suspension are related to cost safety; therefore, budgets, monitoring mechanisms, and shutdown conditions should be established during testing.

Advantages and limitations

Main advantages

• Multiple types of open-source applications can be launched in the cloud, reducing the need for local installation, GPU configuration, and troubleshooting of dependencies.
• It also offers packages based on duration or number of uses, as well as APIs for use cases that require a customized experience, infrequent usage, or system integration.
• Team permissions, point tracking, and automatic suspension help multiple users manage the costs of cloud applications together.

Current restrictions

• The old SD model set can no longer be used through the original entry point; although there are still historical explanations in the documentation, these can easily lead to confusion with the current available state.
• Cloud applications are composed of multiple projects and models, and their quality, account management methods, data rules, and maintenance status vary.
• The rules regarding privacy, services, and refunds are not clearly defined; it is necessary to get confirmation from the platform before handling sensitive data or making large deposits.
• The models and prices listed in the API documentation may be outdated; therefore, the production system must query these details in real time and implement failover mechanisms in case of issues.

Frequently Asked Questions

Can the StarHan Future SD model set still be used directly?

It is not possible to use the old entry point for confirmation at the moment. The old SD section will redirect to the exploration page, and the list of public applications does not show the SD version; it is necessary to search for and check the real-time details first.

Is the platform free?

Browsing is available for free, but cloud applications are charged based on the duration of their use; packages are billed according to the number of calls made, and public model APIs are also charged on a per-use basis.

How much is 1 point worth?

According to the help center, 1 yuan allows the purchase of 1 point. Different applications consume different numbers of points every 10 minutes; the example values given are between 10 and 80 points.

What is the difference between pay-per-use packages and cloud application credits?

The pay-per-use plan charges based on the number of times the large model is invoked, while cloud-deployed applications are billed according to their runtime. The applications covered by these two approaches differ, as do the conditions that trigger billing.

Is an API provided?

Available. The platform offers interfaces for public chat completion, as well as lists of vectors and models; once a user logs in, an API Key is generated, and the actual models must first be verified through the list interface.

Is the entire platform open source?

No. Only some of the cloud-native projects developed by official organizations are made public and distributed under the Apache-2.0 license; this does not mean that all AI application platforms, models, and third-party applications are open source.

Can points or packages be refunded?

No unified refund policy can be found on the current public page. Before making a top-up, it is necessary to confirm with the platform regarding unused points, the number of available packages, and the procedures for dealing with application errors.

Summary

The former SD model set from StarHan Future has evolved into a comprehensive AI application marketplace; it is now more suitable for experiencing various applications in the cloud, purchasing pay-per-use packages, or invoking multiple model APIs. Before using it, it is necessary to check whether the old SD capabilities are once again available, to understand the three different billing methods, and to verify each aspect such as the status of the applications, licenses, data rules, and refund conditions.