Seedream API Explained: The Text Engine for AI Workflows

What is the Seedream API?

The Seedream API is a hosted, uncensored large language model (LLM) service designed specifically for text generation tasks within creative and technical pipelines. Unlike general-purpose chat interfaces, this API is engineered to handle the nuanced demands of prompt engineering, captioning, and metadata extraction without the typical content refusals found in standard models. It serves as the intelligent text engine that powers the input and output stages of AI workflows, particularly those involving visual media.

At its core, the API uses an open-weight model tuned for adult and creative content, ensuring that your specific vision isn't filtered out by safety rails designed for broader audiences. It delivers text in, text out, supporting streaming via Server-Sent Events (SSE) and tool/function calling. This makes it ideal for developers who need reliable, deterministic text outputs to drive downstream processes, such as generating prompts for image generators or scripting voiceovers.

Text vs. Image: Understanding the Roles

Many developers confuse the role of text models with image generation models. The Seedream API does not generate images, audio, or video; it generates the text instructions that drive those models. In a typical Stable Diffusion workflow, the text model is responsible for interpreting user intent, expanding brief ideas into detailed prompts, and refining those prompts based on feedback.

• Prompt Engineering: The API expands simple concepts like "cyberpunk city" into detailed prompts with lighting, camera angle, and style descriptors.
• Post-Processing: It can analyze generated images (via external OCR or vision models) and generate captions or metadata.
• Scripting: For video pipelines, it generates scene descriptions and dialogue scripts.

This separation of concerns allows you to choose the best image model while keeping your text generation flexible and uncensored. The Seedream API acts as the brain of the operation, while image models act as the hands.

Why Uncensored Text Matters

Standard LLMs often apply content filters that can interfere with creative or adult-themed work. For example, a model might refuse to describe a "naked statue" because it detects "naked" and assumes a sexual context, even if the intent is artistic. The Seedream API is tuned to answer without these refusals for lawful adult use, ensuring your creative vision isn't arbitrarily cut.

This uncensored nature is crucial for prompt generation. If the text model refuses to generate a detailed description of a mature scene, the resulting image will be missing key details. By removing these barriers, you gain full control over the content. Note that there is one hard limit: sexual content involving minors is always blocked, regardless of context. This ensures the API remains broadly usable while maintaining a baseline of safety.

For security researchers or writers exploring controversial topics, this API provides a reliable text engine that doesn't second-guess your intent. It’s a pragmatic tool for professionals who want their text output to reflect their input exactly, without corporate moderation filters.

Use Case: Prompt Generation for SD

One of the most common uses for the Seedream API is generating high-quality prompts for Stable Diffusion. Stable Diffusion models respond best to structured, descriptive text. The API can take a rough idea and expand it into a detailed prompt with specific keywords for style, lighting, and composition.

For example, you can send a prompt like "a futuristic car" and receive a detailed output: "A sleek, neon-lit futuristic sports car, cyberpunk aesthetic, rain-slicked streets, volumetric lighting, 8k, highly detailed, unreal engine 5 render." This level of detail significantly improves the quality of the generated image.

The API supports tool/function calling, allowing you to structure the output JSON for direct ingestion by your image generation pipeline. This reduces the need for post-processing text parsing, making your workflow faster and more reliable. Whether you’re generating thousands of variations or just a few, the uncensored model ensures that your specific stylistic choices aren’t filtered out.

Use Case: Metadata & Captioning

Beyond generating prompts, the Seedream API is excellent for post-processing text tasks. If you have a large dataset of images or videos, you can use the API to generate accurate captions and metadata. This is particularly useful for building searchable libraries or training custom models.

For instance, you can provide a description of an image’s content, and the API will generate a concise, keyword-rich caption suitable for SEO or database indexing. The 100,000-token context window allows you to process large blocks of text or multiple images’ worth of metadata in a single request, reducing latency and cost.

This capability extends to scriptwriting for video pipelines. The API can generate scene descriptions, character dialogues, and voiceover scripts, all tailored to your specific creative direction. The uncensored nature ensures that mature themes in your scripts aren’t flagged as errors, allowing for more authentic storytelling.

Integration with OpenAI SDKs

The Seedream API is fully OpenAI-compatible, meaning you can use it with any library or tool that supports the OpenAI API format. This includes the official Python SDK, JavaScript SDK, and many third-party tools. You simply need to update the base URL and API key in your configuration.

• Base URL: https://api.sdxlapi.com/v1
• Model ID: uncensored
• Endpoint: POST /v1/chat/completions

This compatibility allows you to swap out your text model without rewriting your entire pipeline. If you’re currently using a different LLM for prompt generation, you can switch to the Seedream API with minimal code changes. The API also supports streaming via SSE, which is useful for real-time applications where you want to display text as it’s generated.

Function calling is also supported, allowing you to define custom tools that the model can use to fetch data or perform actions. This makes the API versatile for complex workflows that require dynamic data integration.

Performance and Context Window

The Seedream API features a 100,000-token context window, which is significantly larger than many standard models. This allows you to process large documents, long scripts, or extensive image metadata in a single request. For example, you can feed a 50-page script into the API and have it generate character bios, scene summaries, and dialogue adjustments all at once.

The API supports up to 300 requests per minute per key, which is sufficient for most batch processing tasks. The maximum request body size is 8 MB, ensuring that you can send large payloads without hitting limits. For most use cases, this provides ample capacity for both real-time and batch operations.

Performance is optimized for text generation, with low latency for simple prompts and slightly longer times for complex, multi-step reasoning tasks. The uncensored model is tuned for speed and accuracy, ensuring that your pipeline doesn’t bottleneck on text generation. If you need higher throughput, you can generate multiple API keys and distribute requests across them.

Pricing and Scalability

The Seedream API uses a transparent, pay-as-you-go pricing model with no monthly fees or subscriptions. The cost is $0.25 per 1 million input tokens and $1.00 per 1 million output tokens. This means you only pay for the text you send and receive, making it cost-effective for both small projects and large-scale pipelines.

For example, generating 10,000 detailed image prompts would cost approximately $0.01 in input and $0.05 in output, totaling $0.06. This low cost allows you to experiment freely with different prompt structures and styles without worrying about high bills.

You can top up your account with as little as $10, with bonuses for larger deposits: +5% bonus credit for $50 and +10% for $100. Paid credit never expires, so you can build up a balance and use it over time. New accounts receive $0.50 in trial credit valid for 7 days, with no credit card required, allowing you to test the API before committing.

Getting Started with Your First Request

Getting started with the Seedream API is straightforward. First, sign up for an account on the "Get API key" page using just an email and password. No phone number or credit card is required for the trial. Once signed up, your API key is displayed immediately.

Next, configure your OpenAI-compatible client with the base URL https://api.sdxlapi.com/v1 and your API key. You can then send a request to the /v1/chat/completions endpoint with the model ID "uncensored." Here’s a simple example using the Python SDK:

The API supports streaming, function calling, and standard chat completions, making it easy to integrate into any workflow. You can regenerate your API key at any time, which revokes the old key, ensuring security. Prompts are not used for training, and your account privacy is protected with just an email and password.