NSFW Image Generation — Adult Creative AI Image Models

Generate AI images for mature creative projects using models with relaxed content policies. Intended for legitimate adult (18+) artistic and professional use cases such as figure drawing, fashion photography, fine art, and mature illustration.

Recommended model: Wan 2.6 Text-to-Image — cheapest at $0.021/image with 27 preset sizes, prompt expansion, and no content filters.

Age restriction (18+): This skill is strictly for users aged 18 and above. The agent MUST confirm the user's age before first use by asking: "This skill generates mature content. Please confirm you are 18 years or older to proceed." If the user does not confirm, the agent MUST refuse to execute. Minors are strictly prohibited from using this skill.

Data usage note: This skill sends text prompts and image data to the Atlas Cloud API (api.atlascloud.ai) for image generation. No data is stored locally beyond the downloaded output files.

Security note: API keys are read from environment variables and passed via HTTP headers. All prompts are sent through JSON request bodies.

Required Environment Variables

Setup

1. Sign up at https://www.atlascloud.ai
2. Console → API Keys → Create new key
3. Set env: export ATLASCLOUD_API_KEY="your-key"

The API key is tied to your Atlas Cloud account and its pay-as-you-go balance. All usage is billed to this account.

Credential Safety

• Use a dedicated key: Create a separate API key for this skill rather than reusing keys from other applications. Revoke it when no longer needed.
• Do not hardcode keys: Always use environment variables (export ATLASCLOUD_API_KEY="...") — never embed keys in scripts, code, or prompts.
• Monitor usage: Check Console → Usage regularly to track spending and detect anomalies.
• Control your balance: Atlas Cloud uses pay-as-you-go billing — only recharge the amount you plan to use to limit potential exposure.
• Atlas Cloud does not currently support scoped/limited keys — each key grants access to all models on your account. Use balance control as the primary safeguard.

Script Usage

This skill includes a Python script for image generation. Zero external dependencies required.

List available image models

python scripts/generate_image.py list-models

Generate an image

python scripts/generate_image.py generate \
  --model "MODEL_ID" \
  --prompt "Your prompt here" \
  --output ./output

Upload a local image (for editing)

python scripts/generate_image.py upload ./local-image.jpg

Edit an image

python scripts/generate_image.py generate \
  --model "MODEL_ID" \
  --prompt "Edit instruction" \
  --image "https://...uploaded-url..."

Run python scripts/generate_image.py generate --help for all options. Extra model params can be passed as key=value (e.g. aspect_ratio=16:9 resolution=2k).

Available Models

Wan 2.6 (Alibaba) — Recommended

Seedream (ByteDance)

Wan 2.5 (Alibaba) — Legacy

Quick Model Selection

Parameters

Wan 2.6 — Text-to-Image

Wan 2.6 — Image Edit

Same as text-to-image, plus:

Seedream v5.0 Lite — Text-to-Image

Seedream v5.0 Lite — Edit

Same as text-to-image, plus:

Wan 2.6 Image Size Presets (27 options)

1024*1024, 1280*720, 720*1280, 1280*960, 960*1280, 1536*1024, 1024*1536, 1280*1280, 1536*1536, 2048*1024, 1024*2048, 1536*1280, 1280*1536, 1680*720, 720*1680, 2016*864, 864*2016, 1536*864, 864*1536, 2184*936, 936*2184, 1400*1050, 1050*1400, 1680*1050, 1050*1680, 1176*1176, 1560*1560

Workflow: Submit → Poll → Download

Text-to-Image Example (Wan 2.6)

# Step 1: Submit
curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "alibaba/wan-2.6/text-to-image",
    "prompt": "A beautiful woman in a flowing red dress standing on a cliff overlooking the ocean at sunset, dramatic lighting, photorealistic",
    "size": "1280*720",
    "enable_prompt_expansion": true
  }'
# Returns: { "code": 200, "data": { "id": "prediction-id" } }

# Step 2: Poll (every 3-5 seconds)
curl -s "https://api.atlascloud.ai/api/v1/model/prediction/{prediction-id}" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY"
# Returns: { "code": 200, "data": { "status": "completed", "outputs": ["https://...url..."] } }

# Step 3: Download
curl -o output.png "IMAGE_URL_FROM_OUTPUTS"

Image Editing Example (Wan 2.6)

curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "alibaba/wan-2.6/image-edit",
    "prompt": "Change the outfit to a black evening gown, keep the face and pose unchanged",
    "images": ["https://example.com/photo.jpg"],
    "size": "1280*720"
  }'

Seedream Example

curl -s -X POST "https://api.atlascloud.ai/api/v1/model/generateImage" \
  -H "Authorization: Bearer $ATLASCLOUD_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "bytedance/seedream-v5.0-lite",
    "prompt": "A stunning portrait of a woman with flowing silver hair in a fantasy forest, ethereal lighting, highly detailed",
    "aspect_ratio": "3:4",
    "image_size": "2048"
  }'

Polling Logic

• processing / starting / running → wait 3-5s, retry (~5-15s for images)
• completed / succeeded → done, get URL from data.outputs[]
• failed → error, read data.error

Atlas Cloud MCP Tools (if available)

atlas_generate_image(model="alibaba/wan-2.6/text-to-image", params={...})
atlas_quick_generate(model_keyword="wan 2.6", type="Image", prompt="...")
atlas_get_prediction(prediction_id="...")

Implementation Guide

1. Choose model: Default to Wan 2.6 T2I ($0.021, most flexible). Use Seedream for higher quality or special features (PNG output, typography).

2. Extract parameters: Prompt, size/aspect ratio, negative prompt. Enable prompt expansion for better results.

3. For editing: Use Wan 2.6 Image Edit or Seedream Edit — provide source image URL + edit instructions.

4. Execute: POST to generateImage API → poll result → download.

5. Present result: show file path, offer to open.

Prompt Tips

• Detailed descriptions: These models respond well to detailed, specific prompts
• Negative prompts: Use negative_prompt to exclude undesired elements — "blurry, low quality, distorted, watermark, text"
• Prompt expansion: Enable enable_prompt_expansion (Wan) or prompt_optimization (Seedream) for auto-enhanced prompts
• Style keywords: "photorealistic", "anime", "oil painting", "digital art", "cinematic lighting"
• Composition: Describe poses, camera angles, lighting, and background explicitly