How can we help you?

Popular searchesPopular questionsCategory navigationPopular articles

Popular articles

📄 API Access Documentation

If you want to integrate GUID.AI's chat, image, video, audio, and other capabilities into your own website, backend, scripts, workflows, or third-party tools, we recommend using the current system's **V1 Open API**, which is called uniformly through `/v1/*` paths. API integration can be understood as 5 simple steps: 1. Log in to your account and create an API Key 2. Confirm that your account balance or quota is normal 3. Call `/v1/models` to view models available to the current key 4. Call text, image, video, or audio APIs based on your scenario 5. Debug and launch based on returned results and error messages --- ### 1. What Do You Need Before API Integration? Before starting integration, we recommend confirming the following items: | Preparation Item | Description | | --- | --- | | Platform account | You need to log in before creating an API Key | | API Key | Authentication credential for calling `/v1/*` APIs, usually starting with `sk-` | | Available balance or quota | API calls and web creation usually share the same account balance | | Server-side environment | Store the API Key in a backend, script, or secure environment variable | | Target model | Visible models may differ between keys, so check the model list first | > Reminder: API Keys should not be placed in frontend pages, public repositories, screenshots, chat records, or browser local scripts. --- ### 2. Step 1: Obtain an API Key In the current system, an API Key may also be displayed as **API key**, **token**, or **Token**. How to obtain it: 1. Log in to your account. 2. Go to the top **Skills / MCP Access** page. 3. Open the **Go to Key Management / API Token Management** page. 4. Click **Create Key**. 5. Fill in the key name and configure as needed: - Call quota - Expiration time - Model restrictions - IP whitelist 6. After creation succeeds, copy and save the full key. Put it uniformly in the request header: ```http Authorization: Bearer sk-your-api-key ``` --- ### 3. Step 2: Confirm Base URL and Authentication Method The unified rules for the current system's public API are as follows: | Item | Description | | --- | --- | | Base URL | Your gateway domain, such as `https://your-domain.com` | | API prefix | `/v1/*` | | Authentication method | `Authorization: Bearer sk-xxx` | | Common request format | `application/json` | | Streaming response | Text APIs support `text/event-stream` (SSE) | That means actual call addresses usually look like this: ```text https://your-domain.com/v1/chat/completions https://your-domain.com/v1/images/generations https://your-domain.com/v1/videos ``` --- ### 4. Step 3: Call the Model List API First Before integrating business APIs, the most recommended first call is: ```bash curl https://your-domain.com/v1/models \ -H "Authorization: Bearer sk-your-api-key" ``` This API is used to: 1. Verify whether the API Key is valid 2. Confirm which models the current key can see 3. Confirm what capabilities each model supports After success, it usually returns: - `success: true` - `data[]` model list - Each model's `id` - Each model's supported `supported_endpoint_types` We recommend using the `data[].id` returned here for the `model` field in later requests. --- ### 5. Step 4: Choose APIs by Scenario The current system's public capabilities are mainly divided into 5 categories: | Scenario | Method | Path | Description | | --- | --- | --- | --- | | Model list | `GET` | `/v1/models` | Get models available to the current key | | Text chat | `POST` | `/v1/chat/completions` | Most commonly used, suitable for chat, writing, Q&A, and tool calls | | Text completion | `POST` | `/v1/completions` | Compatible with the older prompt completion method | | Responses conversation | `POST` | `/v1/responses` | Multi-turn, structured output, and reasoning scenarios | | Image generation | `POST` | `/v1/images/generations` | Text-to-image | | Image editing | `POST` | `/v1/images/edits` | Image editing and image-to-image | | Video generation | `POST` | `/v1/videos` | Create asynchronous video tasks | | Video query | `GET` | `/v1/videos/{task_id}` | Query task status | | Video download | `GET` | `/v1/videos/{task_id}/content` | Download video or cover | | Multimodal video | `POST` | `/v1/guidai/videos` | Enhanced modes such as image-to-video, first-and-last frames, and continuation | | Speech synthesis | `POST` | `/v1/audio/speech` | Text-to-speech | | Voice list | `GET` | `/v1/audio/voice/{model}` | Query available voices for a TTS model | If this is your first integration, we recommend the following order: 1. `/v1/models` 2. `/v1/chat/completions` 3. `/v1/images/generations` 4. `/v1/videos` 5. `/v1/audio/speech` --- ### 6. Most Common Integration Examples #### 1. Text Chat This is the most commonly used API and is recommended first for new projects. ```bash curl https://your-domain.com/v1/chat/completions \ -H "Authorization: Bearer sk-your-api-key" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-4o", "messages": [ {"role": "user", "content": "Help me write a product introduction"} ] }' ``` Suitable scenarios: - AI chat - Copywriting - Coding assistance - Structured output - Multi-model application wrappers #### 2. Image Generation ```bash curl https://your-domain.com/v1/images/generations \ -H "Authorization: Bearer sk-your-api-key" \ -H "Content-Type: application/json" \ -d '{ "model": "dall-e-3", "prompt": "A cat sitting by the window, morning light, cinematic feel", "size": "1024x1024", "response_format": "url" }' ``` Suitable scenarios: - Text-to-image - Poster generation - Product image creation - Visual asset production #### 3. Video Generation Video APIs are usually **asynchronous tasks** and require 3 steps: ```bash # 1. Create a task curl https://your-domain.com/v1/videos \ -H "Authorization: Bearer sk-your-api-key" \ -H "Content-Type: application/json" \ -d '{ "model": "sora-2", "prompt": "A cat walking by the sea at sunset, cinematic camera style", "size": "1280x720", "seconds": 8 }' ``` ```bash # 2. Query status curl https://your-domain.com/v1/videos/task_xxx \ -H "Authorization: Bearer sk-your-api-key" ``` ```bash # 3. Download result curl -L https://your-domain.com/v1/videos/task_xxx/content \ -H "Authorization: Bearer sk-your-api-key" \ -o output.mp4 ``` #### 4. Text-to-Speech We recommend querying voices first, then starting speech synthesis: ```bash curl https://your-domain.com/v1/audio/voice/tts-1 \ -H "Authorization: Bearer sk-your-api-key" ``` ```bash curl https://your-domain.com/v1/audio/speech \ -H "Authorization: Bearer sk-your-api-key" \ -H "Content-Type: application/json" \ -d '{ "model": "tts-1", "input": "Welcome to the speech synthesis service", "voice": "alloy", "response_format": "mp3" }' \ -o speech.mp3 ``` --- ### 7. Which Text API Should You Use? Many developers hesitate between the 3 text APIs during first integration. You can choose like this: | API | When to Use | | --- | --- | | `/v1/chat/completions` | Use it by default for most new projects | | `/v1/completions` | Use it when you need compatibility with older systems or older SDKs | | `/v1/responses` | Use it when you need multi-turn context, reasoning, structured output, or enhanced tool calling | If you are not sure yet, start directly with `/v1/chat/completions`. It is usually the safest choice. --- ### 8. Common Response Formats During Integration #### 1. Text APIs They usually return JSON, where the most common fields are: - `choices` - `message` - `content` - `usage` #### 2. Streaming Text APIs If you pass `stream: true`, the response is usually `text/event-stream`, which is an SSE stream. #### 3. Image APIs They usually return generated result information and the corresponding `usage`. #### 4. Video APIs When creating a task, they first return: - `id` - `status` - `progress` Then poll the status through the query API. #### 5. Audio APIs TTS usually returns a binary audio stream directly, not JSON. --- ### 9. What Should You Know About Billing and Quota? In the current system, API calls usually check: 1. Whether the API Key is valid 2. Whether the user status is normal 3. Whether the balance or quota is sufficient 4. Whether an IP whitelist restriction is hit 5. Whether model restrictions or quota restrictions for the key are exceeded Also note: - Visible models may differ between API Keys - Different models may use different billing methods - Text is usually billed by tokens - Images may be billed per generation - Videos may be billed per generation or by duration If you are launching in production, we recommend: - Split keys by environment - Set quota limits for test keys - Set an IP whitelist for production keys - Regularly check wallet balance and consumption records --- ### 10. How Do You Troubleshoot Common Errors? | Symptom | Common Cause | Recommended Solution | | --- | --- | --- | | `401` | API Key is wrong, Bearer is missing, or the key is invalid | Check the `Authorization` request header and key status | | `403` | Insufficient permission, IP whitelist restriction, or unavailable model | Check IP, model permissions, and account status | | `400` | Missing parameters or wrong format | Check fields such as `model`, `messages`, `prompt`, and `voice` | | `404` | Wrong path or task does not exist | Check the URL and whether `task_id` is correct | | `429` | Upstream load is high or requests are too frequent | Reduce frequency and retry later | | `500` | Upstream or gateway internal exception | Record request parameters and response content, then retry later or contact the administrator | You can troubleshoot in this order: 1. Check whether the key is correct first 2. Then check the Base URL and API path 3. Then check whether `model` comes from `/v1/models` 4. Then check balance, quota, and IP whitelist 5. Finally check the request body field format --- ### 11. Security Recommendations Before Launch To avoid API Key leakage or misuse, we recommend: 1. Do not hard-code API Keys in frontend code 2. Store them only in backend environment variables or configuration centers 3. Create different keys for development, testing, and production 4. Set different quota limits for different projects 5. Enable IP whitelist for production keys 6. Rotate long-term keys regularly 7. If leakage is suspected, immediately delete the old key and recreate one --- ### 12. Recommended Integration Path If you are a backend developer, we recommend this path: ```text Create an API Key | ├─ Call /v1/models ├─ Select a text model ├─ Run /v1/chat/completions successfully ├─ Integrate image / video / audio as needed └─ Then implement quota limits, whitelists, exception retries, and logging ``` If you are building an Agent or automation workflow, you can further integrate Skills or MCP after running the API successfully. --- ### Frequently Asked Questions | Question | Answer | | --- | --- | | Are API Key and login session the same thing? | No. When calling `/v1/*` public APIs, use an API Key instead of a web login session. | | Are API and web creation balances the same? | They usually share the same account balance system. | | Why can I enter the API page but calls still fail? | The key may be invalid, quota may be insufficient, an IP whitelist may restrict it, model permissions may not match, or request body fields may be wrong. | | Why does the video API not directly return an mp4? | Because video generation is usually asynchronous. Create a task first, query status, then download the result. | | Which API should I integrate first? | Start with `GET /v1/models`, then `POST /v1/chat/completions`. | | Why do I still get an error after passing a model name? | The model may not be visible to the current key. Check `/v1/models` first. | | How do I enable streaming output? | Pass `stream: true` in the text request body. The server usually returns SSE. | | Why does the audio API not return JSON? | Because TTS returns a binary audio stream. | --- ### One-Sentence Summary In the current system, the safest API integration path is: **create an API Key first, use `/v1/models` to confirm available models, then run the first call through `/v1/chat/completions`, and gradually integrate image, video, and audio capabilities.**

GUID.AI Help Center

Read

📄 Creator Monetization Guide

If you want to turn your creative capabilities on GUID.AI into continuous income, the current system is best suited for these 4 paths: **publish paid inspirations, accumulate work assets, invite friends to receive computing power rewards, and use API / Skills to provide external services**. Simple explanation: | Monetization Method | Suitable For | Revenue Source | | --- | --- | --- | | Publish paid inspirations | Image, video, and audio creators | Other users purchase your work inspirations | | Accumulate work assets | Creators with continuous output capability | Increase repeat purchases and exposure through serialized publishing | | Invite friends for computing power rewards | People with communities, private traffic, or content distribution capability | Rebates after invited users register or recharge | | API / Skills external services | Developers, studios, and operations teams | Package platform capabilities into your own products or services | --- ### 1. The Most Direct Monetization Method: Publish Paid Inspirations This is the clearest and most direct creator monetization method in the current system. You can first generate works through **Image / Video / Audio / Apps**, then publish satisfactory results to the **Inspiration Square** and set them as **paid inspirations**. After other users purchase them, the system automatically handles deduction, revenue crediting, work copy creation, and related flows. Under the current rules, after creators publish paid inspirations: 1. The work must be set to **public** so others can see it. 2. The work can be set as **free** or **paid**. 3. If it is set as paid, a single-use price must be filled in. 4. After other users purchase successfully, the revenue enters your **wallet income details**. Suitable content types include: - High-quality prompt works - Reusable image style templates - Video scripts or shot plans with clear scenarios - Dubbing, narration, or voice sample examples - Finished materials suitable for derivative creation > Recommendation: Do not publish only works that "look good". Prioritize works that "others can pick up and continue using", because this type of content is more likely to convert. --- ### 2. Recommended Monetization Loop: Create -> Publish -> Earn -> Scale If you are just starting as a creator, we recommend following this path: 1. Enter **Chat / AI Image / AI Video / AI Audio / Apps** and produce content first. 2. Save the result to **My Works**. 3. In **Assets -> My Works**, organize the title, tags, description, and prompt. 4. Select high-quality works and publish them to the **Inspiration Square**. 5. First test clicks and favorites with free works, then gradually list paid works. 6. After revenue enters **My Wallet**, regularly review which content converts best. 7. Turn well-performing themes into series to continuously increase repeat purchases and exposure. The core of this method is not "publish one item and make money", but to turn works into reusable, discoverable, and continuously accumulating assets. --- ### 3. What Makes a Work Easier to Sell? In the current system, whether users are willing to purchase an inspiration usually depends on 4 factors: #### 1. Whether the Title Is Clear Do not make the title too abstract. Try to let people understand at a glance what the work can do. For example: - Instead of "Futuristic Poster" - Prefer "Cyberpunk New Product Launch Poster Prompt" #### 2. Whether the Tags Are Accurate Tags affect how efficiently works are discovered in the Inspiration Square. We recommend covering: - Content type: image / video / audio - Style type: realistic, illustration, cinematic, Chinese style, anime - Use scenario: e-commerce, short video, promotional poster, brand display #### 3. Whether the Prompt or Description Has Practical Value If your work only displays the result, users may favorite it. But if you clearly explain the prompt idea, parameter direction, and usage suggestions, users are more willing to pay for it. #### 4. Whether Pricing Is Reasonable At the beginning, it is not recommended to price too high. You can first test market feedback with a lower threshold, then gradually adjust the price based on favorites, conversion, and repeat purchases. > Pricing too low is not helpful for building perceived value, while pricing too high affects first purchases. A more stable approach is to get transactions working first, then gradually raise the price. --- ### 4. Besides Selling Inspirations, How Else Can You Monetize? #### 1. Use Invite Friends for Distribution-Based Monetization If you already have an official account, community, social circle, short video account, knowledge community, or a group of users willing to try AI tools, **Invite Friends** is a very suitable second growth curve. The current system supports: - Generate invitation links - Generate invitation QR codes - Download invitation posters - View invitation records - View invitation reward rules Recommended approach: 1. First publish your own works on social media to attract user attention. 2. Then guide users to register on the platform through invitation links or posters. 3. When users recharge or continue using the platform later, you can receive rebates according to page rules. This method is suitable for people with traffic distribution capability, especially: - AI tutorial bloggers - Community operators - Design service providers - Short video content creators - Enterprise training or consultant accounts #### 2. Use the API to Build "Your Own Product" If you are a developer, or you have a team that can package simple products, the platform API can help you upgrade from "selling individual works" to "selling tool capabilities". You can build on the API: - Enterprise internal AI content tools - Official account auto-generation assistants - Image-text / video production tools - Private workflows or automation scripts - Customer-facing web products, mini programs, or SaaS The revenue from this model no longer depends only on individual works, but comes from: - Project delivery fees - Membership fees - Per-call charges - Enterprise service fees #### 3. Use Skills / MCP for Agent Scenario Services If you work on Agents, office automation, intelligent assistants, or creative workflows, you can integrate platform capabilities through **Skills / MCP** and further turn them into your own solutions. Suitable for: - Delivering Agent workbenches to customers - Building creative production pipelines for teams - Creating intelligent assistants that can write, draw, and make videos This path is more like professional service monetization, and the average order value is usually higher than simply selling inspirations. --- ### 5. Where Can You View Revenue After It Arrives? Whether it is inspiration revenue sharing or invitation rebates, we recommend managing everything centrally in **My Wallet**. In the wallet, you can usually view: - Current available balance - Cumulative income - Income details - Consumption records - Withdrawal entry If your revenue has reached the platform requirements, you can initiate a withdrawal in the wallet. According to existing documentation, **withdrawable balance usually comes from creator revenue sharing or rebates**; balance obtained through recharge is usually used for consumption and is not withdrawable. > Reminder: Whether withdrawal is available, the minimum withdrawal amount, and the arrival time are subject to the real-time rules on the wallet page and withdrawal popup. --- ### 6. Which Monetization Method Fits Different Creator Types? | Creator Type | More Suitable Path | | --- | --- | | People who can write prompts and create images | Start with paid inspirations | | People who can create video packaging or shot scripts | Create serialized video inspirations | | People who can dub or create audio samples | Create reusable audio templates | | People with communities and traffic | Invitation rebates + inspiration distribution | | People who can develop tools | API / Skills service monetization | | People with studios or teams | Run both inspiration sales and project delivery | If you currently have no traffic and no technical team, the most recommended starting point is **paid inspirations**, because it has the lowest threshold and is the easiest monetization method to start with in the current system. --- ### 7. The Most Stable Beginner Starting Plan If you want to complete your first income as soon as possible, follow this sequence: 1. Choose the direction you are best at first: image, video, or audio. 2. Create 10 to 20 high-quality works in a row. 3. Pick 3 to 5 works with the highest reuse value and complete their titles, tags, and prompts. 4. Publish 1 to 2 free works first for exposure. 5. Then publish 2 to 3 low-threshold paid works to test conversion. 6. Check Inspiration Square feedback and wallet income details once a day. 7. Continue turning themes that convert well into series. 8. Share these works on social media and add invitation links to amplify rebates. This approach covers: - Content production - Work exposure - First transaction - Revenue review - Scaling --- ### Frequently Asked Questions | Question | Answer | | --- | --- | | Can I monetize even if I have no followers now? | Yes. Start by publishing paid inspirations and focus on works with reuse value instead of pursuing external traffic first. | | Do I have to know development to monetize? | No. You can monetize through inspiration publishing and invitation rebates without development skills. If you can develop, you can further scale revenue with API / Skills. | | Which method is best for beginners? | Paid inspirations are best for beginners because the path is shortest and the startup cost is lowest. | | Why did I publish a work but get no revenue? | Common reasons include insufficient exposure, unclear title and tags, unsuitable pricing, or weak reuse value. | | Can revenue be withdrawn? | Creator revenue sharing and rebates are usually withdrawable balance, but whether they can be withdrawn, how to withdraw, and the minimum amount are subject to wallet page rules. | | Can recharged balance be withdrawn? | Usually no. Recharged balance is mainly used for creation, purchasing inspirations, or API calls. The withdrawable portion usually comes from revenue sharing or rebates. | | Is there any value in creating free works? | Yes. Free works help you accumulate exposure, favorites, and trust, which can then drive later paid work conversions. | | Should I create inspirations first or invite users first? | If you have creation capability, start with inspirations. If you have traffic capability, adding invitation rebates will be more effective. | | Who are API and Skills more suitable for? | They are more suitable for developers, teams, studios, or people who want to package platform capabilities into services. | --- ### One-Sentence Summary In the current system, the most realistic creator monetization path is: **create content assets first, then publish paid inspirations; if you have traffic, add invitation rebates; if you have technical capability, upgrade to API / Skills service monetization.**

GUID.AI Help Center

Read

📄 Complete tutorial on multi-model collaboration

`Multi-model Collaboration` is an advanced Q&A mode on the GUID.AI chat page. After it is enabled, the system calls multiple participating models **in parallel** for the same question, lets each model answer separately, and then uses a **summary model** to aggregate the results into a more complete conclusion that is better suited for decision-making. It is especially suitable for the following scenarios: - Complex question verification - Multi-solution comparison - High-value content creation - Professional cross-validation - Tasks that need to balance quality, speed, and cost Compared with normal single-model chat, multi-model collaboration provides **broader perspectives, higher fault tolerance, and more stable conclusions**. Note that it usually also brings **higher consumption** and **longer response time**. --- ### 1. What Is Multi-model Collaboration? In normal chat, a question is usually answered by only one model. In multi-model collaboration, one question goes through the following process: 1. You select **2 to 5 participating models** 2. The system sends the same question to these models at the same time 3. Each model independently outputs its own answer 4. After all participating models finish answering, the system calls **1 summary model** 5. The summary model summarizes, compares, and refines the answers, then outputs the final collaborative conclusion Simple explanation: - **Participating models** are responsible for "thinking separately" - **Summary model** is responsible for "integrating everything" This mechanism significantly reduces the risk that a single model misses key information, misunderstands the question, or produces a one-sided answer. --- ### 2. Where Do I Enter It? 1. Enter the platform **Chat page**. 2. In the toolbar above the input box at the bottom of the page, find the **`Multi-model Collaboration`** button. 3. Click it to open the **Multi-model Collaboration configuration panel**. Before enabling, the button usually appears in a normal state. After enabling, it displays **`Multi-model Collaboration · Enabled`** and shows the current collaboration status near the toolbar. --- ### 3. Complete Usage Steps #### Step 1: Select Participating Models In the **`Participating Models`** area of the configuration panel, check the models you want to participate in answering. Recommended rules in the current system: - **Select at least 2** - **Select at most 5** Selection suggestions: - For more stable comprehensive judgment: select 3 to 4 models with different styles - To control cost: select 2 core models - For in-depth comparison: mix reasoning, knowledge-oriented, and speed-oriented models Common pairing ideas: - **Reasoning + general-purpose**: suitable for solution analysis, code ideas, and complex Q&A - **General-purpose + fast**: suitable for high-frequency business Q&A - **Mixed vendors**: suitable for cross-validation and reducing single-model bias > Tip: The more participating models you choose, the more comprehensive the conclusion usually is, but consumption also increases accordingly. #### Step 2: Select the Summary Model In the **`Summary Model`** dropdown, select one model responsible for integrating the final result. The summary model's job is not to answer the question again, but to: - Summarize the views of participating models - Identify consensus and differences - Filter duplicate content - Provide the final recommendation or conclusion Prefer models with: - Strong reasoning ability - Good long-text integration ability - High stability If the participating models already include a high-quality model, you can usually let one of them also serve as the summary model. #### Step 3: Set the Scheduling Strategy The current system supports the following scheduling strategies: | Scheduling Strategy | Suitable Scenarios | Characteristics | | --- | --- | --- | | `Speed First` | Pursuing faster results | More suitable for daily efficiency-oriented Q&A | | `Price First` | Wanting to control consumption | More suitable for batch use or cost-sensitive scenarios | | `Success Rate First` | Pursuing stable completion | More suitable for important tasks, formal output, and complex questions | How to choose: - Daily office work and general consultation: use `Speed First` - High-frequency use and team cost control: use `Price First` - Formal proposals, important analysis, and customer delivery: use `Success Rate First` #### Step 4: Configure Web Search You can configure the **`Web Search`** switch in the collaboration panel. Recommended to enable for: - Latest information - Public information references - Supplementing time-sensitive content Recommended to disable for: - Pure internal knowledge organization - Fixed copy generation - Tasks more sensitive to cost and speed If web search is also enabled on the chat page itself, it usually participates in the overall collaboration process as well. #### Step 5: Send the Question and View the Results After collaboration is enabled, enter your question in the chat input box and send it. The system automatically performs the following process: 1. **Parallel answering**: multiple participating models start answering at the same time 2. **Streaming display**: each model's output is displayed in real time as an independent message card 3. **Automatic aggregation**: after participating models finish, the summary model outputs the comprehensive result You will usually see two types of messages: - Original answers from each participating model - One final **collaboration summary** message --- ### 4. Recommended Question Style To make multi-model collaboration truly valuable, we recommend asking questions as clearly as possible: #### 1. Clear Goal Do not simply ask "What should I do?" Instead, try to clearly state: - What your goal is - What the background is - Who the result is for - What output format you want Example: ```text Please analyze the plan for adding multi-model collaboration capability to an existing system from the perspectives of product, technical implementation, and cost control, and provide a recommended solution with risk notes. ``` #### 2. Specify Comparison Dimensions If you want model outputs to be more comparable, you can directly specify dimensions in the question, such as: - Accuracy - Cost - Implementation complexity - User experience - Delivery timeline #### 3. Specify Output Format For example, require: - Table - Bullet-point summary - Risk list - Execution steps - Final recommendation This makes it easier for the summary model to produce structured results. --- ### 5. Which Scenarios Are Best Suited? | Scenario | Recommendation Level | Reason | | --- | --- | --- | | Technical solution design | ✅ Strongly recommended | Suitable for multi-perspective analysis of architecture, implementation path, risks, and cost | | Business decision comparison | ✅ Strongly recommended | Suitable for comparing pros and cons of multiple options and forming comprehensive recommendations | | High-value copywriting or proposals | ✅ Recommended | Multiple models can provide different expression ideas, then the summary model polishes them uniformly | | Fact-checking and professional Q&A | ✅ Recommended | Helps cross-validate and reduce single-model bias | | Casual chat | ❌ Not recommended | Higher cost than single-model chat and not cost-effective | | Scenarios that strongly require second-level response | ❌ Not recommended | Requires waiting for multiple models and summary flow, so the response is slower | --- ### 6. How to Combine Models More Effectively? #### Combination 1: Stable - Participating models: 2 to 3 - Summary model: 1 strong reasoning model - Scheduling strategy: `Success Rate First` - Suitable for: formal replies, customer proposals, important reports #### Combination 2: Efficient - Participating models: 2 - Summary model: 1 general-purpose model - Scheduling strategy: `Speed First` - Suitable for: daily office work, quick comparison, lightweight analysis #### Combination 3: Cost-Controlled - Participating models: 2 - Summary model: 1 economical or general-purpose model - Scheduling strategy: `Price First` - Suitable for: high-frequency use, budget-sensitive teams #### Combination 4: Deep Research - Participating models: 4 to 5 models from different vendors - Summary model: 1 strong integration model - Scheduling strategy: `Success Rate First` - Web search: recommended to enable - Suitable for: industry research, competitor analysis, complex decisions --- ### 7. Relationship with Other Chat Capabilities Multi-model collaboration is not an isolated feature. It can work with other capabilities on the chat page: | Capability | Can Be Combined | Description | | --- | --- | --- | | `Web Search` | Yes | Suitable for questions that need time-sensitive information | | `Deep Thinking` | Yes | Suitable for more complex and rigorous reasoning tasks | | `Role` | Can be combined based on page capabilities | Makes answers better fit a position or tone requirement | | `Preset Templates` | Can be combined based on page capabilities | Suitable for fixed-format tasks such as summarization, analysis, and writing | Recommended order: 1. Clarify the task goal first 2. Select participating models and a summary model 3. Enable web search or deep thinking as needed 4. Send the question last --- ### 8. How Is Consumption Calculated? This is the most important point to understand before using multi-model collaboration. Multi-model collaboration is not "one request, multiple answers"; it is **the accumulation of multiple model calls**: - Each participating model is billed independently - The summary model is also billed independently - Enabling web search, deep thinking, or longer output may further increase total consumption Example: - Select `3 participating models` - Select `1 summary model` Then one question is actually equivalent to at least **4 model calls**. Therefore, we recommend: - Use single-model mode first for simple questions - Enable collaboration for important questions - Reduce the number of participating models when controlling cost - Before batch use, test the best configuration with a small number of samples ---

GUID.AI Help Center

Read

Frequently asked questions

📄 How do I use the Skills feature?

In the current system, **Skills** are an **Agent skill package integration method** built on top of Guid AI's open capabilities. They are suitable for directly calling the platform's **image generation, video generation, and audio generation** capabilities inside Agent conversation environments such as **Claude, Cursor, and Codex**, without manually writing HTTP API requests. The current official skill package name is: `guid-ai-skills` > Note: If you want to integrate capabilities into your own website, backend system, app, or automation process, use the **V1 Open API** (`/v1/*`) first. If you want to integrate with a client that supports MCP, you can also choose the **MCP service** (`/mcp`). Skills are more suitable for scenarios where you want to use multimodal capabilities directly inside Agent conversations. --- ### 1. What Can Skills Do? Skills in the current system mainly cover the following capabilities: | Capability Category | Supported | Description | | --- | --- | --- | | Model list query | ✅ | View currently available models | | Text-to-image | ✅ | Start image generation directly in an Agent conversation | | Standard video generation | ✅ | Supports task creation, polling until completion, and result download | | guidai multimodal video | ✅ | Supports video generation scenarios with multiple modes | | Voice query / text-to-speech | ✅ | Call TTS in a conversation | | Text conversation API | ❌ | Skills do not provide Chat / Completions / Responses. The conversation itself is provided by the Agent | In one sentence: **Skills help the Agent call the platform's image, video, and audio capabilities, but they do not replace the chat model itself.** --- ### 2. Which Scenarios Are Suitable for Skills? We recommend using Skills in the following scenarios: 1. You usually work in Agents such as **Claude, Cursor, or Codex**, and want to generate images, videos, or speech directly through natural language. 2. You do not want to study integration details such as API parameters, signatures, polling, and file downloads yourself. 3. You want to use content generation capabilities as part of an Agent workflow. If you have the following needs, Skills are not recommended as the first choice: - Backend system integration, batch calls, or business API development: use the **V1 Open API** - Integration with MCP-capable clients using a standard protocol: use the **MCP service** - Calling text conversation capabilities: use the Agent itself directly, or call `/v1/chat/completions` --- ### 3. Skills Usage Flow #### Step 1: Prepare an API Key First create an available API Key in the platform backend. The format is usually: ```text sk-xxxxxxxx ``` After creation, save it securely. Skills will use this key later to call platform capabilities. --- #### Step 2: Obtain and Install the Skill Package The skill package name used by the current system is: ```text guid-ai-skills ``` Installation method: 1. Obtain the `guid-ai-skills.zip` package. 2. Extract it into the Skills directory of the Agent you use. 3. After extraction, the directory should contain at least the `SKILL.md` file. Example directory structure: ```text guid-ai-skills/ ├─ SKILL.md ├─ .env └─ ... ``` Common notes: | Platform | Installation Notes | | --- | --- | | Claude-like Agents | Place it in the Skills directory supported by that client | | Cursor | Place it in a project-level or user-level Skills directory, depending on the client version | | Codex-like Agents | Place it in the corresponding Skills / agent skills directory | | Other Agents | Refer to each platform's documentation for Skills installation | > Skills directory locations vary between clients. If you cannot find the directory, check the official instructions for the corresponding Agent first. --- #### Step 3: Configure Environment Variables Fill in the following configuration in `guid-ai-skills/.env`, or configure them as system environment variables: ```bash GUIDAI_BASE_URL=https://guidai.lychessclub.work GUIDAI_API_KEY=sk-xxx ``` Field descriptions: | Variable Name | Required | Description | | --- | --- | --- | | `GUIDAI_BASE_URL` | Yes | Guid AI gateway address. Do not include path suffixes such as `/v1` | | `GUIDAI_API_KEY` | Yes | The API Key you created on the platform | After configuration is complete, we recommend closing and reopening the Agent to ensure the skill package loads the latest environment variables. --- #### Step 4: Use Directly in Conversation After installation and configuration are complete, you can directly describe your requirement in an Agent conversation. The Agent will automatically call the corresponding capability according to the workflow defined in `SKILL.md`. For example, you can directly enter: - `List the currently available image models for me` - `Generate an illustration of a sunset beach with a cinematic feel, 16:9` - `Generate an 8-second video based on this reference image` - `Convert this copy into a female voice audio` If configured correctly, the Agent will automatically complete model query, task submission, polling, or result retrieval, without requiring you to manually assemble API requests. --- ### 4. How Should You Choose Between Skills, API, and MCP? | Method | Suitable For | Description | | --- | --- | --- | | `Skills` | Agent users such as Claude, Cursor, and Codex | Suitable for directly calling image, video, and audio capabilities in chat conversations | | `V1 Open API` | Developers, backends, and third-party systems | Suitable for system integration, batch calls, and API development | | `MCP` | Clients that support the MCP protocol | Suitable for integration through the standard tool protocol | Simple decision guide: - Want to generate images, videos, or speech in one sentence inside an Agent: use **Skills** - Want to write code to integrate with the platform yourself: use the **V1 Open API** - Want to use the MCP standard protocol: use **MCP** --- ### 5. Common Troubleshooting | Issue | Possible Cause | Solution | | --- | --- | --- | | Skills are not visible after installation | The skill package directory is wrong, or the client has not reloaded | Check whether it has been extracted to the correct Skills directory and restart the Agent | | Authentication fails during calls | `GUIDAI_API_KEY` is missing, incorrect, or expired | Recheck the API Key in `.env` and confirm the key status is normal | | Wrong call address / cannot connect | `GUIDAI_BASE_URL` is configured incorrectly | Make sure the gateway root address is filled in without extra paths | | Models can be queried but content cannot be generated | The current key has restricted permissions, insufficient quota, or the model is unavailable | Check token quota, model binding restrictions, and upstream model status | | Want to use Skills for text conversation | Skills themselves do not cover text conversation APIs | For text Q&A, use the Agent directly or call `/v1/chat/completions` | | Are Skills and MCP the same thing? | No | Skills are a skill package method, while MCP is a standard protocol integration method | --- ### 6. Recommended Usage Suggestions 1. We recommend creating a separate API Key for Skills to make usage tracking and permission isolation easier. 2. If you often do multimodal creation inside Agents, use Skills first because the integration threshold is the lowest. 3. If you later need to connect the same capabilities to a business system, switch to the `/v1/*` Open API after becoming familiar with the capabilities. 4. If your team already has an MCP tool system, you can further evaluate whether to integrate through `/mcp`. ---

GUID.AI Help Center

Read

Frequently asked questions

📄 How do I switch models?

In `GUID.AI`, different feature pages usually support selecting models separately. The most common way to switch models is: **enter the corresponding page, click the "Model" dropdown on the page, and then select the model you want to use from the list**. After switching successfully, newly submitted conversations or generation tasks will use the new model. ## 1. Which Pages Support Model Switching? In the current system, the following pages usually support model switching: - **Chat page**: used for text Q&A, writing, translation, summarization, coding assistance, and more - **Image page**: used for text-to-image, reference image generation, image editing, and more - **Video page**: used for text-to-video, image-to-video, and more - **Audio page**: used for text-to-speech, dubbing, and more Notes: - Model lists on different pages are independent of each other - The model you select on the chat page will not automatically sync to image, video, or audio pages - Some models appear only on specific capability pages. For example, image models do not appear in the chat model list ## 2. How Do I Switch Models on the Chat Page? 1. Enter the **Chat page**. 2. Find the current model name or **"Please select a model"** near the input area. 3. Click the model selector to open the available model list. 4. Select the model you want to use. 5. Return to the input box and send a message. New questions will be answered using the model you just selected. Recommended choices: - Daily Q&A: prioritize the default or recommended model - Faster results: choose a faster-response model - Lower consumption: choose a more economical model - Complex analysis: choose a model with stronger reasoning ability Tips: - If **Multi-model Collaboration** is enabled, the system calls multiple models according to the collaboration configuration instead of using only a single model - After switching models, we recommend experiencing the difference starting from the next new message ## 3. How Do I Switch Models on the Image Page? 1. Enter the **AI Image** page. 2. Find **Model Selection** above the generation area or in the parameter area. 3. Click the dropdown to view the image model list. 4. After selecting the target model, fill in the prompt, upload reference images, and click **Generate**. Notes: - Different image models may support different capabilities, such as reference images, aspect ratios, resolution, and quality parameters - After switching models, some parameter options on the page may change accordingly - If a model does not support the feature you are currently using, try switching to another model ## 4. How Do I Switch Models on the Video Page? 1. Enter the **AI Video** page. 2. Find the **Model** dropdown in the video generation form. 3. Select the video model you want to use. 4. Continue filling in the prompt, uploading materials, and submitting the task. Notes: - Different video models may support different resolutions, aspect ratios, subtitles, audio, lip sync, and other capabilities - After switching models, we recommend confirming generation parameters again to avoid mismatches between parameters and model capabilities ## 5. How Do I Switch Models on the Audio Page? 1. Enter the **AI Audio** page. 2. Find the **Select Model** dropdown in the audio generation area. 3. Click and select the audio model you need. 4. Then select a voice, enter text, and start generating. Notes: - Some voices or sound parameters are associated with the current model's capabilities - If a voice becomes unavailable after switching models, it is usually because the model does not support the current voice or configuration ## 6. Why Cannot I See Any Switchable Models? If you do not see a model dropdown, common causes include: - The current page does not support manual model switching - The current feature exposes only one available model - Account permissions, plan, or available capability scope are limited - The current model list is still loading - Upstream service exceptions prevent the model list from returning normally You can try: - Refreshing the page and checking again - Switching to pages that support model selection, such as chat, image, video, or audio - Checking account balance, permissions, or plan scope ## 7. Why Cannot Some Pages Switch Models? This is normal system design. For example, **Apps** tools are usually pre-bound by operations to a fixed **app + channel + model**, and frontend users generally **cannot freely switch models** when using them. This is designed to ensure: - More stable results - Easier parameter adaptation - More controllable task routing - Simpler user operations If you want to switch models freely, we recommend using native creation pages such as chat, image, video, and audio first. ## 8. What Should I Pay Attention to After Switching Models? - **Switching only takes effect for subsequent new tasks** and will not automatically rerun previously submitted tasks - **Different models have different consumption**, so check the estimated consumption or balance changes on the page before generation - **Different models produce different effects**, and the same prompt may produce very different results under different models - **Some models have different capabilities**, so parameter changes after switching are normal ## Frequently Asked Questions | Question | Explanation | | --- | --- | | Why do results change a lot after switching models? | Different models have different styles, capabilities, speeds, and costs, so output differences are normal. | | Why is the model I want not in the model list? | The model may not be available on the current page, or it may not be available for the current account. | | Will previous content be automatically regenerated after switching models? | No. Switching affects only newly submitted requests. | | Why do some parameters disappear after switching? | It means the new model does not support some previous capabilities or parameter configurations. | | Why cannot I switch models on the Apps page? | Apps generally run with fixed bound models, and frontend users usually cannot change them manually. | ## One-Sentence Summary When you want to switch models, enter the corresponding **Chat / Image / Video / Audio** page, find the **Model dropdown**, and select again. If a page has no switching entry, it usually means the feature has a fixed model bound in the system and does not support free frontend switching.

GUID.AI Help Center

Read

Frequently asked questions

📄 How do I register a GUID.AI account?

Registering a GUID.AI account usually takes only a few minutes. You can register with a **username and password**, or directly log in/register with **Google or other third-party accounts** when the system has enabled the corresponding capability. --- ### 1. Go to the Registration Page 1. Open the official GUID.AI website. 2. Click **"Register / Log In"** in the upper-right corner of the page. --- ### 2. Register with Username and Password If the page provides a standard registration method, follow these steps: 1. Enter your **username**. 2. Enter your login password. 3. Enter the password again to confirm it. 4. If the page requires email verification, fill in your email address and get a verification code. 5. Enter the email verification code you received. 6. Read and check the user agreement, privacy policy, and related items if required by the page. 7. Click the **"Log In / Register"** button to submit. 8. If this is your first time using the system, logging in will also complete registration. --- ### 3. Register Quickly with a Third-Party Account If the registration or login page displays third-party login buttons, you can also register directly with those accounts: - Google - Other third-party login methods configured by the site The process is as follows: 1. Click the corresponding third-party login button on the login/registration page. 2. Go to the third-party platform and complete authorization. 3. After authorization succeeds, the system automatically creates a GUID.AI account for you, or logs you in to an existing account. This method is suitable for users who do not want to remember a separate username or receive verification codes. --- ### 4. If the Page Requires an Email Verification Code 1. First enter an email address that can receive messages normally. 2. Click **"Get Verification Code"**. 3. Go to your mailbox and check the verification code email. 4. Return to the page, enter the verification code, and then submit registration. If you do not receive the email for a long time, we recommend: 1. Check the spam folder. 2. Confirm whether the email address was entered correctly. 3. Wait a moment and request the verification code again. --- ### 5. If Human Verification Appears on the Page The current system supports Turnstile human verification. If the page indicates that the environment is being verified, please wait a few seconds before continuing to submit registration. If human verification is not completed, the system may temporarily prevent sending verification codes or submitting registration. --- ### 6. What Can You Do After Successful Registration? After completing registration and logging in, you can usually use the following features immediately: 1. Enter AI creation pages such as chat, image, video, and audio. 2. View the Inspiration Square and work content. 3. Enter the Wallet page to view your balance or recharge. 4. Create API keys and access open APIs. 5. Complete account information in the personal center, bind an email address, or bind third-party accounts. If the platform has enabled new-user rewards, you may also automatically receive a certain amount of initial computing power after successful registration. The specific details are subject to the page display and site rules. ---

GUID.AI Help Center

Read

Still need help?

Our support team is ready to provide personalized assistance anytime.