Prefect Pony XL is a free, polished Pony-based anime checkpoint with deep artist comprehension and strong NSFW knowledge. Use the Pony score ladder, Euler a, 25 to 30 steps, CFG 5 to 7 (lower on merged versions), Clip skip 2, an external SDXL VAE, and 832×1216. We tested it on an RTX 4090 and a 3060.

Prefect Pony XL NSFW Guide: Setup and Settings (2026)

Prefect Pony XL is one of the cleaner, more refined Pony Diffusion fine-tunes you can run for anime NSFW work in 2026. It keeps Pony’s signature deep artist knowledge and strong anatomy while delivering a more polished default aesthetic than raw Pony V6. If you want to preview the look before downloading, you can try our on-site generator first as a no-install option.

We ran Prefect Pony XL through our standard prompt battery. This guide covers setup, the settings that work, and how to prompt it correctly with Pony score tags. Since it shares the Pony lineage, read it alongside our AutismMix guide and our anime checkpoint roundup.

What Prefect Pony XL is

Prefect Pony XL is a fine-tuned and merged checkpoint built on Pony Diffusion V6 XL. The author has iterated it across many versions, refining anatomy, reducing the chaos that raw Pony is known for, and tightening the default look toward a clean anime style. It keeps Pony’s enormous artist comprehension, which is the main reason people choose the Pony lineage in the first place.

It is fully uncensored, so explicit content is part of its native vocabulary with no unlock required. The official model lives on its Civitai page, where the latest version sits alongside the earlier releases.

A key difference from the Illustrious-lineage models like WAI and NoobAI: Prefect Pony uses Pony score tags, not Illustrious quality tags, and it typically wants an external SDXL VAE loaded.

It is also worth knowing where this checkpoint sits in the wider landscape. Pony Diffusion V6 became the dominant base for anime and NSFW work because of how deeply it learned artist styles and explicit concepts, but its raw output could be inconsistent and demanding to prompt. A whole generation of fine-tunes, Prefect Pony among the most polished, exists to keep that comprehension while smoothing the rough edges. Choosing Prefect Pony means you want the Pony brain with a calmer, more finished default look, and you are happy to handle the two or three Pony-specific setup steps that come with the territory.

Hardware and software

We benchmarked Prefect Pony XL on two GPUs.

On an RTX 4090 (24 GB) an 832×1216 image at 28 steps renders in about 4 to 6 seconds. With hires fix, figure 12 to 16 seconds for a finished image. On an RTX 3060 (12 GB) the base image takes roughly 18 to 25 seconds, and a hires-fixed final lands around 45 to 60 seconds. As an SDXL-class model, 8 GB is the realistic floor and 12 GB is comfortable.

For software, any modern UI works with no special prediction mode:

• Automatic1111 for the standard interface.
• Forge for speed and memory efficiency.
• ComfyUI for node graphs, batching, and detailer chains.

If you would rather not install anything, the no-install generator is the quick path to a single image.

Recommended settings for Prefect Pony XL

These are the settings we landed on after testing. Note that recommended CFG depends on whether you are using a standard or merged version.

• Sampler: Euler a is the safe default. DPM++ 2M Karras gives sharper output.
• Steps: 25 to 30 for standard versions. Some merged versions run well at 10 to 20 steps.
• CFG scale: 5 to 7 for standard versions. Merged versions often want a much lower CFG of around 2, so check the version notes.
• Resolution: 832×1216 portrait, 1216×832 landscape, or 1024×1024.
• Clip skip: 2 (the author uses Clip skip 2 on every image).
• VAE: load an external sdxl_vae.safetensors. Pony-lineage models often render washed-out colors without it.

For finishing, the author’s own workflow leans on hires fix or img2img at higher resolution, ADetailer for face detail, and tiled diffusion for img2img upscaling. We upscale 1.5x with R-ESRGAN 4x+ Anime6B and denoise 0.35 to 0.45.

Prompting Prefect Pony XL

Because Prefect Pony is Pony-based, you prompt with the Pony score ladder. This is essential, not optional; quality collapses without it. The standard quality block is the full score chain. Keep negatives reasonable rather than enormous.

Here is a starting prompt we used in testing.

Positive:
score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up,
source_anime, BREAK,
1girl, solo, long hair, looking at viewer, detailed face,
nude, large breasts, lying on bed, indoors, soft lighting,
detailed background, depth of field

Negative:
score_6, score_5, score_4, worst quality, low quality, lowres,
bad anatomy, bad hands, missing fingers, extra digits,
watermark, signature, blurry, censored

Notes from our runs:

• Always lead with the full score ladder. On Pony-lineage models this chain is load-bearing.
• Use “source_anime” to steer the aesthetic; “source_pony” or “source_furry” shift it.
• The BREAK keyword separates the quality block from the content block so the model weights them independently.
• Artist tags work strongly thanks to Pony’s deep artist comprehension. Use them ethically and respect living artists.
• Keep negatives reasonable. You can put lower score tags like score_4 in the negative to push quality up.

Prefect Pony XL versus the other anime NSFW checkpoints

Here is how Prefect Pony compares against the models we benchmark, on a single 832×1216 image at 28 steps on the RTX 4090.

Checkpoint	Base	Quality tags	Best sampler	Steps	CFG	4090 gen time
Prefect Pony XL	Pony	score_9, score_8_up…	Euler a	25-30	5-7 (2 on merges)	~5s
AutismMix Confetti	Pony	score_9, score_8_up…	Euler a	25-30	5-7	~5s
WAI-NSFW Illustrious	Illustrious	masterpiece, best quality…	Euler a	25-40	5-7	~5s
NoobAI-XL	Illustrious	masterpiece, best quality…	Euler a	25-30	5-6	~5s
Hassaku XL	Illustrious	masterpiece, best quality…	Euler a	20-28	3-7	~4s

The takeaway: Prefect Pony and AutismMix are the two Pony-lineage picks, both with score tags and deep artist comprehension. Prefect Pony leans toward a polished anime look out of the box, while AutismMix Confetti is famous for predictability and lean negatives. If you live in the Pony LoRA ecosystem and want a clean default, Prefect Pony is a strong choice.

A practical Prefect Pony workflow

This is the loop we use for finished images.

1. Confirm an external SDXL VAE is loaded; this is the most common Pony setup miss.
2. Generate at 832×1216, Euler a, 28 steps, CFG 6, Clip skip 2, to lock composition.
3. Fix the seed once the pose and framing are right.
4. Run hires fix at 1.5x, R-ESRGAN Anime6B, denoise 0.4, or use tiled diffusion for img2img upscaling.
5. Run ADetailer on the face at inpaint denoise around 0.3, and on hands if needed.
6. Optional final img2img upscale at denoise 0.2 for 4K.

The VAE check at step one saves a lot of confusion over washed-out colors.

LoRA compatibility

Prefect Pony is Pony-based, so it works with the large Pony LoRA library. A few notes:

• Pony character and concept LoRAs load cleanly; start around weight 0.8.
• Illustrious and NoobAI LoRAs do not transfer, since the base is Pony.
• Stacking many LoRAs degrades coherence; keep total weight under about 1.5.
• Because Prefect Pony already has a polished look, you often need fewer style LoRAs.

If you have a collection of Pony LoRAs, Prefect Pony and AutismMix are the two natural homes for them.

Common problems and fixes

Colors look washed out or gray. You are missing an external VAE. Load sdxl_vae.safetensors. This is the single most common Prefect Pony problem.

Output is generic or low quality. You dropped the score ladder. On Pony-lineage models the score_9, score_8_up chain is essential.

Image looks over-contrasted on a merged version. Merged versions often want a much lower CFG, around 2. Check the specific version notes and lower CFG.

Hands or fingers are wrong. Universal SDXL issue. Use hires fix plus ADetailer with a hand model.

Censoring bars appear. Add “censored, mosaic censoring, bar censor” to the negative.

Faces look low detail at full body. Run ADetailer on the face after hires fix.

Understanding the score tag ladder

The score ladder is the most important concept for any Pony-lineage model, so understand it rather than just pasting it. Pony Diffusion V6, which Prefect Pony is built on, was trained with aesthetic score tags from score_1 to score_9, where score_9 marks the highest-rated images. Putting the full ascending chain in your positive prompt tells the model to draw from its best-quality distribution.

The conventional block, score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, works because each “_up” tag includes that level and everything above it, stacking the quality signal. Dropping part of the chain visibly lowers quality on Prefect Pony. Placing low score tags like score_4 or score_5 in the negative pushes further from mediocre output, a clean way to lift quality without bloating the positive.

Source tags shape the overall aesthetic:

• source_anime steers toward the anime illustration look most readers want here.
• source_pony shifts toward the show-inspired style.
• source_furry and source_cartoon move the look accordingly.

Get the score ladder and a source tag right and you have handled most of what makes Prefect Pony prompting work.

Tag weighting and emphasis

After the score ladder, attention weighting is your steering tool. In Automatic1111 and Forge syntax, wrap a tag in parentheses with a number, like (detailed face:1.2) to strengthen or (background:0.8) to weaken. Prefect Pony responds well, but Pony-lineage models can be more sensitive than Illustrious ones, so stay conservative.

Patterns we use:

• Strengthen the focal subject modestly, for example (long hair:1.1).
• Weaken a dominating element, such as (depth of field:0.8).
• Bump an explicit tag only slightly; over-weighting it distorts anatomy quickly on Pony models.
• Keep individual weights under about 1.25 on Prefect Pony to avoid warping.

Inpainting and editing Prefect Pony output

Running Prefect Pony locally lets you repair one region instead of rerolling. The loop:

1. Send the image to the inpaint tab.
2. Mask only the broken area.
3. Set inpaint area to “only masked” for full resolution on that patch.
4. Use denoise 0.4 to 0.6, and keep the score ladder in the inpaint prompt, since dropping it lowers quality even on a small patch.
5. Confirm your external VAE is still active for the inpaint pass, or colors on the patch can drift.
6. Trim the rest of the prompt to just the masked subject.

Keeping the score tags and the VAE active during inpainting are the two Pony-specific details people miss, and they explain most cases where an inpainted patch comes back looking off.

Running Prefect Pony on lower-end hardware

Below 12 GB of VRAM, Prefect Pony still runs with care. On 8 GB cards, use Forge for its better SDXL memory handling, generate at 832×1216, keep batch size at 1, and run hires fix as a separate img2img pass. If you are using a merged version that supports low steps and a low CFG around 2, that combination is genuinely fast even on modest hardware, which is one reason the merged variants are popular with budget-GPU users.

Below 8 GB, local SDXL gets painful. That is where the on-site generator is the better option, running the compute remotely so an older card or laptop can still produce anime output. We treat it as the zero-setup path and local Prefect Pony as the power-user path.

Why choose Prefect Pony XL

The reason to pick Prefect Pony is simple: you want Pony’s enormous artist comprehension and strong anatomy, but with a cleaner, more polished default than raw Pony V6. It delivers that, works with the entire Pony LoRA ecosystem, and is well behaved once you remember the two Pony-specific setup items, namely the score ladder and the external VAE. We keep it installed next to AutismMix for Pony work, and reach for WAI, NoobAI, or Hassaku when we want the Illustrious lineage instead.

Settings recap and final verdict

All in one place: for standard versions, set Euler a, 25 to 30 steps, CFG 5 to 7, Clip skip 2, 832×1216, and load an external SDXL VAE. For merged versions, check the notes, since many want a much lower CFG around 2 and fewer steps. Always lead with the full score ladder plus source_anime and BREAK, keep negatives reasonable, and finish with hires fix or tiled img2img upscaling plus ADetailer on the face.

Our verdict after testing across versions: Prefect Pony XL is one of the cleanest, most polished Pony-lineage anime checkpoints in 2026. It keeps Pony’s deep artist comprehension and strong anatomy while delivering a more refined default than raw Pony V6. Once you handle the three Pony-specific setup items, the score ladder, Clip skip 2, and the external VAE, it is well behaved and consistent.

The natural alternative is AutismMix, the other refined Pony pick, famous for predictability and lean negatives. For the Illustrious lineage with cleaner default anatomy, WAI, NoobAI, and Hassaku are the options. But if you want a polished Pony look and you live in the Pony LoRA ecosystem, Prefect Pony XL is an easy recommendation.

Download it from the Civitai page, load an external SDXL VAE, set Euler a, 28 steps, CFG 6, Clip skip 2, 832×1216, and lead with the score ladder. And if you just want one image without a GPU, the on-site generator is always available.

Frequently asked questions

What are the best settings for Prefect Pony XL?

For standard versions, use Euler a, 25 to 30 steps, CFG 5 to 7, Clip skip 2, and a resolution of 832×1216. Load an external sdxl_vae.safetensors, since Pony-lineage models often render washed out without it. Merged versions of Prefect Pony often want a much lower CFG around 2 and fewer steps, so always check the specific version notes. For finishing, use hires fix at 1.5x with R-ESRGAN Anime6B and denoise around 0.4.

Why are my Prefect Pony colors washed out?

You are almost certainly missing an external VAE. Pony-lineage models, including Prefect Pony XL, frequently render gray or washed-out colors unless you load an external SDXL VAE such as sdxl_vae.safetensors, available from Civitai or HuggingFace mirrors. Drop it in your VAE folder and select it in your UI. This is the single most common Prefect Pony setup mistake, and fixing it instantly restores proper saturation and contrast.

What quality tags does Prefect Pony XL use?

Because Prefect Pony is built on Pony Diffusion, it uses the Pony score ladder, not Illustrious quality tags. Lead your positive prompt with score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, then source_anime and BREAK. The score chain is load-bearing; quality collapses without it. You can also place lower score tags like score_4 in the negative to push quality up. This is different from WAI, NoobAI, and Hassaku, which use masterpiece, best quality.

What CLIP skip should I use with Prefect Pony XL?

Use Clip skip 2. The author applies Clip skip 2 on every image, and it is the standard for Pony-lineage SDXL checkpoints. Setting Clip skip to 1 will noticeably change and usually degrade the output on these models. Confirm it is set to 2 in your UI settings before generating. Combined with the external VAE and the score ladder, Clip skip 2 is one of the three Pony-specific setup items to get right.

Do Illustrious or NoobAI LoRAs work with Prefect Pony XL?

No. Prefect Pony XL is built on the Pony base, so it works with Pony LoRAs, not Illustrious or NoobAI LoRAs, which use a different conditioning scheme and will not bind correctly. If you have a library of Pony character or concept LoRAs, Prefect Pony is a natural home for them; start around weight 0.8. For Illustrious LoRAs, use an Illustrious checkpoint like WAI, NoobAI, or Hassaku instead.

Is Prefect Pony XL better than AutismMix?

Both are excellent Pony-lineage anime checkpoints that use the same score tags, so it comes down to aesthetic preference and workflow. Prefect Pony leans toward a polished anime look out of the box, while AutismMix Confetti is famous for predictability and lean negatives. We keep both installed and choose by project. If you want the cleanest default Pony look, Prefect Pony is great; if you want the most predictable, low-fuss Pony, AutismMix Confetti is hard to beat.

Can Prefect Pony XL run on an RTX 3060 12GB?

Yes. On our RTX 3060 12GB, an 832×1216 base image at 28 steps took about 18 to 25 seconds, and a hires-fixed final landed around 45 to 60 seconds. The 12 GB of VRAM is comfortable for SDXL-class models including hires fix at 1.5x. Cards with 8 GB can run it with longer waits. For quick one-off images without a GPU at all, the on-site generator is the easier route.

What sampler is best for Prefect Pony XL?

Euler a is the safe default and gives Prefect Pony a clean, soft anime look. If you want sharper edges and more fine detail, DPM++ 2M Karras is a strong alternative. Both pair well with 25 to 30 steps on standard versions. Whichever you pick, keep Clip skip at 2 and load an external SDXL VAE, since the sampler choice matters far less than getting those two Pony-specific settings right.