ADetailer is an AUTOMATIC1111 extension that automatically detects faces (or other areas) in generated images and runs a targeted inpainting pass to improve detail – no manual masking required. Install from the Extensions tab. Use denoising 0.4-0.5 to enhance without changing face identity. Works with all NSFW checkpoints.
ADetailer (After Detailer) solves a persistent problem in AI image generation: faces and fine detail degrade at smaller render sizes. A 512px or even 768px generation often produces soft, inconsistent facial features even when the overall composition is strong. ADetailer fixes this by running an automatic detection-and-inpaint pass on faces (or any detected region) immediately after the main generation, without requiring you to manually mask anything.
This guide covers ADetailer installation, configuration, recommended settings for NSFW content, and how to combine it with Hires.fix and ControlNet for maximum output quality.
Prerequisites
Working AUTOMATIC1111 installation with at least one checkpoint. ADetailer works with any checkpoint including NSFW-trained models – there is nothing content-policy-specific about the extension itself. GPU recommended; it adds a fast inpainting pass after each generation.
Step 1 – Install ADetailer
Navigate to Extensions > Available > Load From. Search for “ADetailer.” Find the extension by Bing-su (the original author) and click Install. After installation completes, go to Installed tab and click “Apply and restart UI.” The ADetailer accordion panel will appear below the main generation settings in both txt2img and img2img. Check the Enable box inside the accordion to activate it per generation.
Step 2 – Choose Your Detection Model
ADetailer ships with several detection models that download automatically on first use. The key options for NSFW generation:
- face_yolov8n.pt – fast, accurate face detection for realistic style images. Best default for photorealistic NSFW checkpoints.
- face_yolov8s.pt – slightly slower but more accurate than yolov8n, better for partially obscured faces.
- anime_face_yolov8.pt – optimized for anime-style facial proportions. Use this for anime checkpoints.
- hand_yolov8n.pt – detects hands. NSFW images with hand interactions benefit from a second ADetailer unit targeting hands.
Step 3 – Configure ADetailer Settings
The critical settings inside the ADetailer accordion:
- Detection confidence: 0.3 default. Lower (0.2) detects more faces including partial/small. Higher (0.5+) only detects high-confidence faces.
- Mask blur: 4-8 pixels. Higher values soften the inpaint edge for more seamless blending.
- Denoising strength: Start at 0.45. This is the most important setting – see Step 4 for detail.
- Inpaint steps: Match your main generation steps (20-28).
- CFG scale: Match or slightly lower than your main CFG (6-7).
- Prompt: Leave blank to inherit main prompt, or write face-specific prompts like
detailed face, beautiful eyes, clear skin.
Step 4 – Tune Denoising Strength
Denoising strength is the single most important ADetailer setting. At 0.3-0.4: subtle enhancement – sharpens detail while fully preserving the original face identity. Best when the face is close to correct and just needs clarity. At 0.45-0.55: moderate regeneration – improves quality significantly and can fix minor proportion issues, small risk of identity shift on low-detail faces. At 0.6+: aggressive regeneration – essentially redraws the face. High quality improvement but the result may not match the original character identity. For most NSFW work, 0.4-0.5 is the right range.
Step 5 – Combine with Hires.fix for Maximum Quality
ADetailer and Hires.fix stack well. The recommended pipeline: generate at 512×768 (or native resolution for your checkpoint), enable Hires.fix at 1.5x with 0.5 denoising for overall sharpness, then enable ADetailer at 0.45 denoising for face/body detail. Hires.fix runs first (improving overall image), then ADetailer runs on the sharpened result. Total generation time increases by 30-60% but the quality improvement is significant for NSFW work where face and body detail matters.
Multiple ADetailer Units
AUTOMATIC1111 supports running multiple ADetailer units in sequence. A common NSFW workflow: Unit 1 with anime_face_yolov8.pt at 0.45 denoising (face detail), Unit 2 with hand_yolov8n.pt at 0.4 denoising (hand correction). Enable Unit 2 by opening the second ADetailer accordion below the first. Each unit runs in sequence, so face improvement happens before the hand correction pass.
Related Guides
ADetailer automates what the manual inpainting workflow does by hand. For pose and anatomy control before ADetailer refinement, see the ControlNet guide. For character consistency across a full image series, read our character consistency techniques guide and the LoRA training guide.
ADetailer Settings Deep Dive
Understanding each ADetailer parameter helps you tune it precisely rather than using it as a black box. Here is what each setting actually controls and what to change for common scenarios.
Detection Confidence (0-1): The minimum confidence score a detected region must hit to trigger inpainting. At 0.3 (default), ADetailer processes most visible faces including partial and angled faces. Raise to 0.5+ if ADetailer is incorrectly triggering on non-face elements (decorative patterns, background objects). Lower to 0.2 if faces in the background are not being detected and processed.
Mask Erosion / Dilation: Positive dilation values expand the mask outward from the detected face boundary. Negative values erode it inward. For face enhancement, slight dilation (4-8px) helps blend the inpaint edge with the surrounding hair and neck. For targeted fixes (eyes only, mouth only), use negative dilation to contract the mask to just the specific facial feature.
Inpaint Padding: Adds context around the masked area during inpainting. Higher padding (32-64px) gives the inpainting model more surrounding context to blend against, which usually improves edge coherence. Lower padding (0-16px) speeds up processing. Start at 32 for best quality.
Use Separate Prompt: When checked, you can write prompts specifically for the ADetailer inpaint pass, separate from the main generation prompt. Use this when you want to give the face inpainting specific instructions not appropriate for the full image prompt – for example, detailed anime face, clear eyes, sharp features as the ADetailer prompt while your main prompt describes the full scene.
ADetailer for Batch Processing
ADetailer runs on every image in a batch generation. If you generate a batch of 8 images with ADetailer enabled, all 8 receive the face enhancement pass. This is the most efficient way to use ADetailer – run your normal batch, get 8 enhanced images rather than 8 rough images requiring manual selection and separate enhancement. The time cost per image is modest: approximately 5-15 seconds per detected face on a mid-range GPU (3080/4070 class).
For production workflows where you need large batches of consistent NSFW character images, the recommended setup is: txt2img with your standard NSFW prompt + ADetailer (anime_face_yolov8 at 0.45 denoising) + Hires.fix (1.5x at 0.4 denoising). This single generation pass produces near-final quality images that require minimal post-processing. The total generation time per image is 60-120 seconds on consumer GPU hardware – comparable to generating and then manually inpainting separately.
Combining ADetailer with LoRA for Character Consistency
ADetailer inherits the active LoRAs from your main generation. If you have a character LoRA active at 0.75 strength in your main prompt, that LoRA is also active during the ADetailer face inpainting pass. This means the character’s face identity is reinforced during enhancement rather than drifting. The result: sharper, more detailed faces that still look like your specific character rather than a generic improved face.
If you want the ADetailer pass to use a different LoRA strength than the main generation (for example, stronger LoRA influence during face enhancement), check “Use Separate LoRA” in the ADetailer settings and set the LoRA strength independently for the inpaint pass. This is a niche setting but useful when character LoRA strength is calibrated for body/costume rendering and you want a slightly different balance for face detail specifically.
For full character consistency methodology including ADetailer integration, see our character consistency techniques guide. ADetailer is one of seven methods covered there for maintaining consistent characters across a series of images.
ADetailer for Specific Body Areas
ADetailer is not limited to faces. With the right detection model, it can automatically detect and enhance any identifiable body area. For NSFW content production, this extends its usefulness well beyond face refinement.
Hand detection (hand_yolov8n.pt): Hand anatomy is one of the most common AI generation failure points. Enabling a second ADetailer unit with hand detection at 0.5 denoising significantly reduces the frequency of malformed finger results. The hand detection model identifies hand regions and runs a corrective inpainting pass with the positive prompt reinforcing correct anatomy: perfect hands, 5 fingers, detailed fingers, correct anatomy. This cannot fix every hand failure (severely distorted hands at 0.5 denoising still appear) but eliminates the minor-to-moderate failures that account for most hand quality issues.
Custom region detection: Community-trained detection models for additional body areas are available on Civitai and Hugging Face. Search for “ADetailer model” on Civitai to find detection models for specific body areas relevant to NSFW content. Place downloaded detection model files in the ADetailer models folder (typically extensions/adetailer/models/) and they appear in the ADetailer model dropdown.
Troubleshooting ADetailer Issues
Common ADetailer problems and their solutions:
ADetailer is not running: The Enable checkbox inside the ADetailer accordion must be checked each session (it does not persist across restarts unless set in default settings). Check that the accordion is expanded and the Enable toggle is on. In AUTOMATIC1111 settings, you can set ADetailer to enable by default under Settings > ADetailer.
Wrong region being detected: ADetailer is detecting a background element as a face. Raise the Detection Confidence threshold from 0.3 to 0.5 or higher to filter out low-confidence detections. Alternatively, mask the generation to keep the main character separate from complex background elements that might confuse the detector.
ADetailer is too slow: Running ADetailer on large batches with high resolution can be slow. Reduce the inpainting resolution to 512px square (sufficient for face enhancement) rather than matching the full image resolution. The ADetailer settings allow separate resolution control for the inpaint pass – use the “Inpaint Width” and “Inpaint Height” settings to cap the inpaint size.
Character face identity keeps changing: Denoising is too high. Drop from 0.5 to 0.38-0.42. At very low denoising (0.35), ADetailer functions more like a sharpening/clarity pass than a regeneration. For maintaining a distinctive character face across many generations, 0.38-0.42 denoising with the character LoRA active is the most reliable setting combination.
For the complete quality pipeline that combines ADetailer with ControlNet and inpainting, see our inpainting guide and ControlNet guide. The three extensions work best in sequence: ControlNet for structural control, ADetailer for automatic enhancement, manual inpainting for any remaining corrections.
ADetailer Quick-Start and Recommended Defaults
Getting ADetailer working with the best default settings for NSFW anime generation: install via Extensions > Available, search “ADetailer,” install, restart. In the ADetailer accordion, check Enable. Detection model: anime_face_yolov8.pt for anime checkpoints, face_yolov8n.pt for realistic checkpoints. Confidence: 0.3. Mask blur: 6. Denoising: 0.45. Leave prompts blank to inherit from main generation. Enable in every generation – ADetailer adds minimal time (5-15s per face) and virtually always improves face quality compared to the same generation without it. For users generating NSFW content with multiple characters per image: enable “Save detected map” in ADetailer settings to see which faces were detected and processed, helping diagnose cases where a secondary character’s face was missed. For further automation, combining ADetailer (face enhancement) with ControlNet (pose control) and inpainting (targeted corrections) creates a near-complete quality pipeline that eliminates most manual post-processing steps. See our ControlNet guide and inpainting guide for those components. All three tools are free extensions within the AUTOMATIC1111 ecosystem.
ADetailer is one of the most impactful extensions available for AUTOMATIC1111, delivering measurable quality improvement with minimal configuration. The defaults described in this guide (anime_face_yolov8.pt, 0.45 denoising, 6px mask blur) work well across the vast majority of NSFW anime generation scenarios. Enable it on every generation session and treat the few seconds of extra processing time as a free quality upgrade. For users who want the complete quality toolkit, ADetailer combined with ControlNet pose control and targeted inpainting covers the three main quality vectors in AI image generation: enhancement (ADetailer), structure (ControlNet), and correction (inpainting).
