wan 2.7 image nsfw
wan 2.7 image nsfw — this page examines the technical documentation and API access for the Wan 2.7 Image model. The cited sources describe model capabilities, API endpoints, and configuration parameters but do not establish an adult-content policy or built-in NSFW filtering. Users should consult current service terms for content guidelines.
Read source
wan 2.7 image nsfw: Model Overview
The Wan 2.7 Image model is a unified visual generation system designed for professional-grade image synthesis. It supports text-to-image, image editing, and image set generation.
Architecture and Design
Wan-Image features a natively unified multi-modal architecture that synergizes the cognitive capabilities of large language models with the high-fidelity pixel synthesis of diffusion transformers. This design seamlessly translates nuanced user intents into precise visual outputs. The system integrates a Planner, an advanced Multi-modal Large Language Model responsible for deep semantic reasoning and task routing, with a Visualizer, a DiT-based module for high-fidelity pixel generation. A four-channel variational autoencoder supports the generation of transparency channels. A Prompt Enhancer boosts logical reasoning to capture nuanced user intent, while an Image Refiner enhances high-frequency details and pushes generation resolution to 4K quality. This architecture acts not just as a rendering engine but as an intelligent design assistant.
Professional Capabilities
Wan 2.7 image nsfw unlocks a suite of professional productivity capabilities including ultra-long text rendering, extreme aspect ratio adaptation up to 1:8, hyper-diverse realistic portrait generation, palette-guided generation with user-specified hex-code palettes, multi-subject identity-preserving generation, logical image series generation of up to 12 images, multi-modal interactive editing, true alpha-channel generation via the 4-channel VAE, and high-efficiency 4K synthesis. Across diverse human evaluations, Wan-Image exceeds Seedream 5.0 Lite and GPT Image 1.5 in overall performance, reaching parity with Nano Banana Pro in challenging tasks. The system revolutionizes visual content creation across e-commerce, entertainment, education, and personal productivity.
wan 2.7 image nsfw: Capabilities and Access
The Wan 2.7 Image model is accessible via API through Alibaba Cloud Model Studio. It offers two model variants: wan2.7-image-pro for high-resolution output and wan2.7-image for faster generation.
API Access and Endpoints
Access requires an API key from Alibaba Cloud Model Studio. The China (Beijing) and Singapore regions have separate API keys and request endpoints that cannot be used interchangeably. Cross-region calls lead to authentication failures or service errors. The HTTP synchronous endpoint returns results in a single request, recommended for most use cases. The HTTP asynchronous endpoint is best for long-running tasks, requiring submission of a task and polling for status and results. The Singapore endpoint is POST https://{WorkspaceId}.ap-southeast-1.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation, and the Beijing endpoint is POST https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/multimodal-generation/generation, replacing {WorkspaceId} with the actual workspace ID.
Supported Tasks and Input Formats
Wan 2.7 image nsfw supports text-to-image, text-to-image-set, image-to-image-set, image editing, and multi-image reference generation. Input images can be provided as publicly accessible URLs (HTTP or HTTPS) or Base64-encoded strings. Supported image formats include JPEG, JPG, PNG (alpha channels not supported), BMP, and WEBP. Image resolution must be between 240 and 8000 pixels with an aspect ratio between 1:8 and 8:1, and file size no more than 20 MB. Up to 9 images can be input simultaneously. The text prompt supports Chinese and English with a maximum of 5000 characters. The model outputs images in PNG format with links valid for 24 hours.
Model Variants and Regional Availability
The Wan 2.7 Image model is available in two variants with different performance characteristics and regional deployment options.
wan2.7-image-pro vs wan2.7-image
The wan2.7-image-pro variant supports 4K output for text-to-image generation, while image editing and image set generation support up to 2K resolution. The wan2.7-image variant offers faster generation but does not support 4K output, only 1K and 2K. Both variants support the same input formats and API endpoints. The pro variant is recommended for high-resolution professional use cases, while the standard variant is suitable for faster, lower-resolution tasks. Both models support parameters such as size, n (number of images), watermark, thinking_mode, color_palette, and seed. The thinking_mode parameter enhances inference capabilities to improve output quality but increases generation time.
Regional Deployment
Wan 2.7 image nsfw is available in two regions: China (Beijing) and Singapore. Each region has separate API keys and request endpoints that cannot be used interchangeably. Cross-region calls result in authentication failures or service errors. Users must ensure they use the correct endpoint and API key for their chosen region. The regional separation allows users to select the deployment location that best suits their latency and compliance requirements. The API documentation provides specific endpoint URLs for each region, and users must replace {WorkspaceId} with their actual workspace ID. Before calling a model, users should check the model list and prices for regional model availability.
API Parameters and Configuration Options
The Wan 2.7 Image API offers extensive configuration parameters for fine-tuning generation output, including resolution, image count, color palette, and thinking mode.
Resolution and Image Count Parameters
The size parameter supports two mutually exclusive methods. Method 1 specifies output image resolution using 1K (1024x1024), 2K (2048x2048), or 4K (4096x4096) specifications. For text-to-image with no image input, all three resolutions are supported. Other scenarios support 1K and 2K. Method 2 allows specifying width and height in pixels, with total pixels between 768x768 and 4096x4096 for text-to-image and between 768x768 and 2048x2048 for other scenarios. The n parameter controls the number of images to generate, ranging from 1 to 4 when image set mode is disabled, and from 1 to 12 when enabled. The actual number of images in set mode is determined by wan 2.7 image nsfw and will not exceed n.
Advanced Configuration Options
The thinking_mode parameter, enabled by default, enhances inference capabilities to improve output quality but increases generation time. It is effective only when image set mode is disabled and no image is input. The color_palette parameter allows custom color themes with 3 to 10 colors, each specified as a hexadecimal color with a percentage ratio summing to 100%. The watermark parameter adds an 'AI Generated' label in the bottom-right corner. The seed parameter allows reproducible outputs using the same random number seed, though image generation is probabilistic and results may vary even with the same seed. The bbox_list parameter enables interactive editing by specifying bounding box coordinates for targeted modifications.
Practical Usage Notes
Effective use of the Wan 2.7 Image API requires understanding of authentication, error handling, and output management.
Authentication and Error Handling
Authentication requires a valid API key from Alibaba Cloud Model Studio, passed in the Authorization header as 'Bearer sk-xxxx'. The Content-Type header must be 'application/json'. For asynchronous calls, the X-DashScope-Async header must be set to 'enable'. Error responses include a code and message field indicating the error cause. Common errors include invalid API keys, invalid parameters, and authentication failures. The API returns a unique request_id for tracing and troubleshooting. Successful task execution returns output with choices containing image URLs, while failed tasks return error codes. Task data is retained for only 24 hours, so generated images should be downloaded promptly.
Output Management and Billing
The API returns image URLs in PNG format that are valid for 24 hours. Users must download generated images promptly as they are automatically purged after this period. The usage object in the response includes image_count, size, input_tokens, output_tokens, and total_tokens, though billing is based on the number of images, not tokens. The n parameter directly affects cost as unit price is multiplied by the number of successfully generated images. Users should confirm model pricing before calling. The response also includes a finish_reason field indicating natural completion ('stop') or other termination reasons. For asynchronous tasks, the task_id is returned for querying status and results.
Frequently Asked Questions
Common questions about the Wan 2.7 Image model and its API.
What should I know about wan 2.7 image nsfw?
The cited technical documentation for Wan 2.7 Image describes model capabilities, API access, and configuration parameters but does not establish an adult-content policy or built-in NSFW filtering. The documentation covers text-to-image, image editing, and image set generation with various parameters for resolution, color palette, and thinking mode. Users should consult current service terms from Alibaba Cloud Model Studio for content guidelines. The API requires authentication and returns error codes for invalid requests, but the sources do not specify content moderation policies. For specific questions about acceptable content, refer to the official service documentation.
What are the system requirements for using the API?
To use the Wan 2.7 Image API, you need an API key from Alibaba Cloud Model Studio, which must be exported as an environment variable. The API supports HTTP synchronous and asynchronous calls. For synchronous calls, use the generation endpoint. For asynchronous calls, use the image-generation endpoint with the X-DashScope-Async header set to 'enable'. The API accepts JSON request bodies with model name, input messages, and optional parameters. Input images can be URLs or Base64-encoded strings. The API returns image URLs in PNG format that are valid for 24 hours.
How does the image set generation feature work?
The image set generation feature allows creating a cohesive set of up to 12 images with consistent visual elements across the series. To enable this mode, set the enable_sequential parameter to true. The n parameter specifies the maximum number of images to generate, with a default of 12. The actual number is determined by wan 2.7 image nsfw and will not exceed n. The model ensures rigorous visual consistency and logical thematic flow across the entire group of samples. This feature is available for both wan2.7-image-pro and wan2.7-image variants, with resolution support up to 2K for image set generation.