IF budget priority → use "fal-ai/flux/dev" (from $0.012/image)
  - Cheapest NSFW image option
  - Set enable_safety_checker=false
  - Good prompt following

IF custom style → use "fal-ai/flux-lora" (from $0.032/image)
  - Custom LoRA support
  - Set enable_safety_checker=false
  - Character consistency, specific aesthetics

IF premium quality → use "seedream/v5" (from $0.032/image)
  - Best image quality
  - Excellent skin tones and photorealism
  - Whitelisted on Atlas Cloud

Text Generation

ALWAYS use DeepSeek V3.2:
  - Model: "deepseek-v3-0324"
  - Endpoint: https://api.atlascloud.ai/v1/chat/completions
  - Price: $0.26/M input tokens, $1.10/M output tokens
  - No content filtering
  - 128K context window

API Call Templates

Video Generation Template (Queue-based)

import requests
import time
import os

API_KEY = os.environ.get("ATLAS_API_KEY")
BASE_URL = "https://queue.atlascloud.ai"

def generate_video(prompt, model_path="fal-ai/wan-spicy/v1", **kwargs):
    """Generate video using Atlas Cloud queue API."""

    payload = {
        "prompt": prompt,
        "negative_prompt": kwargs.get("negative_prompt", "low quality, blurry, distorted, deformed"),
        "num_frames": kwargs.get("num_frames", 81),
        "resolution": kwargs.get("resolution", "720p"),
        "guidance_scale": kwargs.get("guidance_scale", 7.5),
        "num_inference_steps": kwargs.get("num_inference_steps", 30),
    }

    # Add image_url for image-to-video
    if "image_url" in kwargs:
        payload["image_url"] = kwargs["image_url"]

    # Add LoRA configuration
    if "loras" in kwargs:
        payload["loras"] = kwargs["loras"]

    # Submit request
    response = requests.post(
        f"{BASE_URL}/{model_path}",
        headers={
            "Authorization": f"Key {API_KEY}",
            "Content-Type": "application/json"
        },
        json=payload
    )
    response.raise_for_status()
    request_id = response.json()["request_id"]
    print(f"Submitted: {request_id}")

    # Poll for completion
    while True:
        status_response = requests.get(
            f"{BASE_URL}/{model_path}/requests/{request_id}/status",
            headers={"Authorization": f"Key {API_KEY}"}
        )
        status = status_response.json()

        if status["status"] == "COMPLETED":
            break
        elif status["status"] == "FAILED":
            raise Exception(f"Generation failed: {status}")

        print(f"Status: {status['status']}...")
        time.sleep(5)

    # Get result
    result = requests.get(
        f"{BASE_URL}/{model_path}/requests/{request_id}",
        headers={"Authorization": f"Key {API_KEY}"}
    ).json()

    video_url = result["video"]["url"]
    print(f"Video ready: {video_url}")
    return video_url

Image Generation Template (Queue-based)

def generate_image(prompt, model_path="fal-ai/flux/dev", **kwargs):
    """Generate image using Atlas Cloud queue API."""

    payload = {
        "prompt": prompt,
        "image_size": kwargs.get("image_size", "landscape_16_9"),
        "num_inference_steps": kwargs.get("num_inference_steps", 28),
        "guidance_scale": kwargs.get("guidance_scale", 3.5),
        "num_images": kwargs.get("num_images", 1),
        "enable_safety_checker": False,  # Always disabled for NSFW
    }

    # Add LoRA configuration
    if "loras" in kwargs:
        payload["loras"] = kwargs["loras"]

    # Submit request
    response = requests.post(
        f"{BASE_URL}/{model_path}",
        headers={
            "Authorization": f"Key {API_KEY}",
            "Content-Type": "application/json"
        },
        json=payload
    )
    response.raise_for_status()
    request_id = response.json()["request_id"]
    print(f"Submitted: {request_id}")

    # Poll for completion
    while True:
        status_response = requests.get(
            f"{BASE_URL}/{model_path}/requests/{request_id}/status",
            headers={"Authorization": f"Key {API_KEY}"}
        )
        status = status_response.json()

        if status["status"] == "COMPLETED":
            break
        elif status["status"] == "FAILED":
            raise Exception(f"Generation failed: {status}")

        print(f"Status: {status['status']}...")
        time.sleep(3)

    # Get result
    result = requests.get(
        f"{BASE_URL}/{model_path}/requests/{request_id}",
        headers={"Authorization": f"Key {API_KEY}"}
    ).json()

    image_urls = [img["url"] for img in result["images"]]
    print(f"Images ready: {image_urls}")
    return image_urls

Text Generation Template (Chat Completions)

def generate_text(prompt, system_prompt=None, **kwargs):
    """Generate uncensored text using DeepSeek V3.2."""

    messages = []
    if system_prompt:
        messages.append({"role": "system", "content": system_prompt})
    messages.append({"role": "user", "content": prompt})

    response = requests.post(
        "https://api.atlascloud.ai/v1/chat/completions",
        headers={
            "Authorization": f"Bearer {API_KEY}",
            "Content-Type": "application/json"
        },
        json={
            "model": "deepseek-v3-0324",
            "messages": messages,
            "max_tokens": kwargs.get("max_tokens", 4096),
            "temperature": kwargs.get("temperature", 0.8),
        }
    )
    response.raise_for_status()

    content = response.json()["choices"][0]["message"]["content"]
    return content

Budget Optimization Rules

When the user does not specify a model, apply these rules:

1. Default to cheapest unless quality is explicitly requested:

   • Video: Wan 2.2 Spicy ($0.03)
   • Image: Flux Dev ($0.012)
   • Text: DeepSeek V3.2 ($0.26/M input)

2. Upgrade when requested:

   • "high quality" / "best quality" → Kling ($0.204) for video, Seedream ($0.032) for image
   • "premium" / "cinematic" → Seedance ($0.222) for video
   • "fast" → Vidu Q3-Turbo ($0.034) for video

3. Warn about costs when:

   • Batch operations exceed 100 items
   • Premium models are used for prototyping
   • Total estimated cost exceeds $10

4. Suggest alternatives:

   • If user picks Kling for drafts, suggest Wan Spicy for testing first
   • If user generates single images, note that Flux Dev is 63% cheaper than Flux LoRA

Safety Checker Settings

• Flux Dev / Flux Dev LoRA: Always set enable_safety_checker: false
• Wan Spicy: No safety checker parameter needed (natively NSFW)
• Seedance / Kling / Vidu: No safety checker parameter needed (whitelisted on Atlas Cloud)
• Seedream: No safety checker parameter needed (whitelisted on Atlas Cloud)
• DeepSeek: No content filter parameter needed (no filtering applied)

Response Format

When generating content, always provide:

1. The model selected and why
2. The estimated cost
3. The complete, runnable code (Python preferred)
4. The output URL(s) when generation completes
5. Suggestions for quality improvement or cost reduction

Important Notes

• Atlas Cloud API keys start with your account identifier
• Queue-based APIs return a request_id for polling
• Chat completions API follows OpenAI-compatible format
• All generated content URLs are persistent (no expiration)
• Rate limits vary by plan — check Atlas Cloud dashboard
• First recharge includes 25% bonus up to $100 at https://www.atlascloud.ai?ref=JPM683&utm_source=github&utm_campaign=nsfw-ai-skill

Why Atlas Cloud for NSFW?

Most AI platforms block NSFW content outright. The few that allow it charge premium prices and offer limited model selection. Atlas Cloud solves both problems:

• Whitelisted Models — Seedance, Kling, Vidu, and more are specifically enabled for NSFW on Atlas Cloud
• No Safety Checker Overhead — Models like Flux Dev run with enable_safety_checker=false by default
• 80%+ Cost Savings — Compared to fal.ai and other providers
• Exclusive Access — Wan 2.2 Spicy is only available through Atlas Cloud
• Enterprise Security — SOC I & II Certified, HIPAA Compliant, US-based

Features

NSFW Video Generation

• Wan 2.2 Spicy — LoRA fine-tuned model purpose-built for NSFW video, from $0.03/s
• Wan 2.2 Spicy LoRA — Custom LoRA variants for specific styles and aesthetics
• Wan 2.5/2.6 — High-quality uncensored video generation with extended duration support
• Seedance v1.5 Pro — Premium video quality, whitelisted exclusively on Atlas Cloud
• Kling v3.0 Pro — Top-tier video generation, 82% cheaper than fal.ai
• Vidu Q3-Pro / Q3-Turbo — Fast, affordable NSFW video at scale

NSFW Image Generation

• Flux Dev — Industry-standard image generation with enable_safety_checker=false
• Flux Dev LoRA — Custom LoRA models for specific styles, characters, and aesthetics
• Seedream v5.0 — Next-generation image model, whitelisted on Atlas Cloud

Uncensored Text Generation

• DeepSeek V3.2 — Full uncensored text generation with no content filtering
• Story writing, dialogue, scene descriptions, character development
• System prompt support for consistent character voice

Advanced Pipelines

• Image-to-Video — Generate a still image, then animate it into video
• Text-to-Image-to-Video — Full pipeline from text description to final video
• Batch Generation — Generate multiple variations in parallel
• LoRA Chaining — Combine multiple LoRA models for unique styles

Supported Models

Complete Model Reference

| Model | Type | Price | Method | Best For | |:------|:-----|:------|:-------|:---------| | Wan 2.2 Spicy | Video | from $0.03/s | LoRA fine-tuned, native NSFW | Budget NSFW video | | Wan 2.2 Spicy LoRA | Video | from $0.03/s | Custom LoRA variants | Styled NSFW video | | Wan 2.5 | Video | from $0.05/s | Uncensored mode | Mid-range video | | Wan 2.6 | Video | from $0.07/s | Uncensored mode | High-quality video | | Seedance v1.5 Pro | Video | from $0.044/s | Whitelisted on Atlas Cloud | Premium video | | Kling v3.0 Pro | Video | from $0.204/s | Whitelisted on Atlas Cloud | Top-tier video | | Vidu Q3-Pro | Video | from $0.06/s | Whitelisted on Atlas Cloud | Quality + value | | Vidu Q3-Turbo | Video | from $0.034/s | Whitelisted on Atlas Cloud | Fast + cheap |

⚠️ Note: Vidu Q3 models may add mosaic/blur to certain NSFW scenes due to training data limitations. Not guaranteed 100% uncensored. For reliable uncensored output, use Wan 2.2 Spicy ($0.03) or Wan 2.6 ($0.07). | Flux Dev | Image | from $0.012/image | enable_safety_checker=false | Budget NSFW image | | Flux Dev LoRA | Image | from $0.032/image | Custom LoRA + safety off | Styled NSFW image | | Seedream v5.0 | Image | from $0.032/image | Whitelisted on Atlas Cloud | Premium image | | DeepSeek V3.2 | Text | $0.26/M input | No content filtering | Stories & dialogue |

Model Selection Guide

Budget Video     → Wan 2.2 Spicy ($0.03)
Quality Video    → Wan 2.6 ($0.07) or Vidu Q3-Pro ($0.06)*
Premium Video    → Kling v3.0 Pro ($0.204) or Seedance v1.5 Pro ($0.222)
Fast Video       → Vidu Q3-Turbo ($0.034)*

* Vidu Q3 may add mosaic/blur to certain NSFW scenes. Use Wan models for reliable uncensored output.
Budget Image     → Flux Dev ($0.012)
Styled Image     → Flux Dev LoRA ($0.032)
Premium Image    → Seedream v5.0 ($0.032)
Text/Stories     → DeepSeek V3.2 ($0.26/M input tokens)

Price Comparison vs fal.ai

| Model | fal.ai | Atlas Cloud | Savings | |:------|:-------|:------------|:--------| | Wan 2.2 Spicy | N/A (not available) | from $0.03/s | Atlas Exclusive | | Wan 2.2 Spicy LoRA | N/A (not available) | from $0.03/s | Atlas Exclusive | | Wan 2.5 (5sec) | $0.25/req | from $0.05/s | 80% cheaper | | Wan 2.6 (5sec) | $0.35/req | from $0.07/s | 80% cheaper | | Seedance v1.5 Pro | N/A (not available) | from $0.044/s | Atlas Exclusive | | Kling v3.0 (5sec) | $1.12/req | from $0.204/s | 82% cheaper | | Vidu Q3-Pro | N/A (not available) | from $0.06/s | Atlas Exclusive | | Vidu Q3-Turbo | N/A (not available) | from $0.034/s | Atlas Exclusive | | Flux Dev | $0.025/req | from $0.012/image | 52% cheaper | | Flux Dev LoRA | $0.050/req | from $0.032/image | 36% cheaper | | Seedream v5.0 | N/A (not available) | from $0.032/image | Atlas Exclusive |

Bottom line: Atlas Cloud is 36-82% cheaper on shared models, and offers 6+ exclusive models not available anywhere else.

Quick Start

1. Get Your API Key

Sign up at Atlas Cloud and grab your API key from the dashboard.

export ATLAS_API_KEY="your-api-key-here"

2. Install the Skill

Add this skill to your agent:

npx skills add https://github.com/thoughtincode/nsfw-ai-skill

3. Start Generating

Just tell Claude what you want:

Generate an artistic nude video with soft lighting using Wan Spicy

Installation

As an Agent Skill

Works with Claude Code, Cursor, OpenAI Codex, GitHub Copilot, Gemini CLI, Windsurf, OpenCode, Kiro, and 15+ AI coding agents.

# Add the skill to your agent
npx skills add https://github.com/thoughtincode/nsfw-ai-skill

Manual Setup

# Clone the repository
git clone https://github.com/thoughtincode/nsfw-ai-skill.git
cd nsfw-ai-skill

# Set your API key
export ATLAS_API_KEY="your-api-key-here"

# Run examples
python examples/video-generation.py
python examples/image-generation.py
python examples/batch-pipeline.py

Dependencies

pip install requests

No other dependencies required. The skill uses Atlas Cloud's REST API directly.

API Integration Guide

Base URL and Authentication

All Atlas Cloud API requests use the same base pattern:

Base URL: https://queue.atlascloud.ai
Auth Header: Authorization: Key {ATLAS_API_KEY}

The API follows a queue-based pattern:

1. Submit a request → get a request_id
2. Poll for status → get the result when complete

Video Generation

Wan 2.2 Spicy — Budget NSFW Video (from $0.03/s)

cURL:

# Submit video generation request
curl -X POST "https://queue.atlascloud.ai/fal-ai/wan-spicy/v1" \
  -H "Authorization: Key ${ATLAS_API_KEY}" \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "A graceful figure dancing in golden sunset light, flowing silk fabric, artistic cinematography, warm color palette",
    "negative_prompt": "low quality, blurry, distorted, deformed",
    "num_frames": 81,
    "resolution": "720p",
    "guidance_scale": 7.5,
    "num_inference_steps": 30
  }'

# Response: {"request_id": "abc123..."}

# Poll for result
curl -X GET "https://queue.atlascloud.ai/fal-ai/wan-spicy/requests/abc123/status" \
  -H "Authorization: Key ${ATLAS_API_KEY}"

# When complete, get result
curl -X GET "https://queue.atlascloud.ai/fal-ai/wan-spicy/requests/abc123" \
  -H "Authorization: Key ${ATLAS_API_KEY}"

Python:

import requests
import time

API_KEY = "your-api-key"
BASE_URL = "https://queue.atlascloud.ai"

def generate_video_wan_spicy(prompt, negative_prompt="", num_frames=81):
    """Generate NSFW video using Wan 2.2 Spicy model."""

    # Submit request
    response = requests.post(
        f"{BASE_URL}/fal-ai/wan-spicy/v1",
        headers={
            "Authorization": f"Key {API_KEY}",
            "Content-Type": "application/json"
        },
        json={
            "prompt": prompt,
            "negative_prompt": negative_prompt or "low quality, blurry, distorted",
            "num_frames": num_frames,
            "resolution": "720p",
            "guidance_scale": 7.5,