Heirloom Male XL started as a personal SDXL merge I built to solve one annoying problem: photoreal models love drifting back to female defaults, especially in couple / romance prompts. I mainly create male-focused (MM) content — including interspecies/fantasy couples — so I tuned this merge to favor male subjects, while keeping an overall clean, photoreal look.

After testing it a bunch in my own workflow, I decided to share it in case it helps other creators who want a stronger male bias without needing ControlNet every time.

🔥 What’s new in V3

Version 3 uses the same merge ingredients as V2 — but with a big upgrade:

✅ I trained a couples-focused LoRA and merged it into the checkpoint.
That means:

• MM / gay couples generate easier with “normal couple prompts”

• Less fighting to keep both subjects male

• More reliable couple composition (still prompt-dependent, but way smoother)

⚠️ Important: V3 is easier, not “automatic.”
You’ll still get best results if you follow the prompt habits from V1/V2 samples (clear genders, clear roles, clear framing).

🧾 Changelog (V2 → V3)

✅ Couples consistency improved (built-in LoRA merge)
✅ Less drift during soft/romantic aesthetics
✅ Better reliability for MM romance prompts
✅ Still keeps the clean photoreal look + prompt-following improvements from V2

📝 Prompt tips (V3 still likes “good prompting”)

Best starting point: use the exact prompts from the sample images on this model page.
Those are the closest “what this checkpoint likes” baseline.

If the model drifts away from male subjects, add one of these (start light, don’t over-weight):

• two men, two asian men,

• male couple, gay couple

• 2boys

• two male lovers

• handsome men, masculine features

👀 Composition tip (for couples)

If you want the couple framed correctly, say the camera framing clearly:

• close-up portrait

• mid shot, waist-up

• full body

• front view, eye level

🎛️ Recommended settings (Forge Neo baseline)

These are the general settings I recommend (same vibe as your V1/V2 samples):

• WebUI: Forge Neo

• Sampler: DPM++ 2M SDE (recommended) / DPM++ 3M SDE

• Schedule: Karras

• Steps: 40–50

• Resolution: 896 × 1344

• CFG: 2.5 → 6.5 (sweet spot; too high can look “plastic”)

⚙️ Forge Neo “Stable Diffusion” settings

Go to: Settings → Stable Diffusion → SD UNet

Use these:

• Clip Skip: 1

• Emphasis Mode: No norm ✅ (helps avoid some SDXL weirdness)

• [PR] Epsilon Scaling: 1.005

• Token Wrap Length: 26

• Tiling: OFF (unless you specifically need tileable images)

• Random Number Generator: GPU

SDXL-specific fields (leave like this unless you know why you’re changing them):

• [SDXL] Crop-Top Coordinate: 0

• [SDXL] Crop-Left Coordinate: 0

• [SDXL] Low Aesthetic Score: 3

• [SDXL] High Aesthetic Score: 6

🧠 Side note (hard-mode concepts = use guidance tools)

Some concepts are just hard mode with pure prompting — especially:

• fantasy couples

• interspecies anatomy (tails/wings/etc)

• complex poses / multiple characters

If you use ControlNet (Canny + Reference Only) it can massively improve:

• couple posing + composition accuracy

• identity consistency (faces/structure)

• fantasy anatomy details

🎛️ Style LoRAs + consistency

Good news: V2 improved the “style LoRA feels weaker after heavy merges” issue, and V3 keeps that improvement.

If a style LoRA still feels muted, try:

• increase LoRA strength slightly

• lower CFG a bit

• do an img2img style pass at low denoise

🚫 Suggested negatives (starter pack — safer baseline)

Use as a baseline and adjust per workflow:

Negatives:

• anime, toon, 2d, 3d, painting, drawing, illustration

• glitch, deformed, mutated, cross-eyed, disfigured, ugly

• distant subjects, wide shot, long shot (only if you want close shots)

• female features (use this only if you keep getting female drift)

• nudity, explicit (only if you want strictly clean outputs)

📸 Sample image transparency

• Same general generation settings as earlier versions

• Some samples may include ADetailer for faces/details (when noted)

🧩 Prompt helper extension (optional but super useful)

If you struggle with prompts, tags, or consistent workflows, I built a helper extension (it was made for my personal use, sharing in case anyone finds it useful):

Prompt Suite Bridge
🔗 https://github.com/EditwithLak/prompt_suite_bridge.git

What it helps with:

• 🗃️ Vault (tag libraries, packs, base prompts, LoRA/TI registry)

• 🧱 Prompt Composer (autocomplete + structured prompting + GGUF LLM support)

• 🧪 Map Generator (canny/openpose/depth map creation + ControlNet routing)

• 🏷️ Caption Tools (optional caption/tag pipelines via separate env)

⚠️ Real talk: I’m not a GitHub pro 😅
If git install is annoying, just Download ZIP from GitHub, extract it into your WebUI extensions/ folder, then restart — follow the instructions in the repo README.

Version 2

is an upgraded refinement of Version 1, built using the same merge ingredients, but tuned for better prompt-following, stronger male/MM consistency, and cleaner results with less “drift.”
The goal stays the same: male-first photoreal, especially for MM romance/couples and fantasy/interspecies concepts — but now it’s more cooperative.

What’s improved in V2

• Better prompt adherence (less “model does its own thing”)

• MM / gay couples generate easier with normal couple prompts

• Fantasy male subjects are more consistent, including interspecies concepts when prompted

• Keeps the clean photoreal look, but with better control and reliability

🧾 Changelog (V1 → V2)

• ✅ Improved prompt-following / instruction compliance

• ✅ More consistent MM / gay couple generation

• ✅ Better fantasy male consistency

• 🎛️ Improved behavior with style LoRAs (see note below)

• 🧪 V2 sample images include an ADetailer pass (V1 samples were generated without helper tools)

📝 Prompt tips

• Best starting point: check the exact prompts used in the sample images on this model page.
  Those prompts are the closest “what this checkpoint likes” baseline.

🧠 Side note (hard-mode concepts = use guidance tools)

Some concepts are just hard mode with pure prompting — especially:
fantasy couples, interracial, interspecies, complex composition/poses.

If you use ControlNet (Canny + Reference Only) it can drastically improve consistency for:

• couple posing + composition accuracy

• identity consistency (faces/structure)

• fantasy anatomy details (tails, wings, wingspans, etc.)

🎛️ Style LoRAs + style consistency (V2 update)

Good news: in Version 2, the “style LoRA feels weaker after heavier merges” issue is largely improved — at least for the styles/LoRAs I personally use and previously had issues with.

That said, I can’t test every single LoRA or style out there. If a LoRA still feels muted, try:

If a style LoRA feels weak, try:

• Increase LoRA strength slightly

• Lower CFG a bit

• Use img2img with low denoise for a style pass

⚙️ CFG Scale tips (recommended)

Don’t crank CFG too high — it can make outputs look over-saturated / plastic.
Don’t go too low either — you’ll lose prompt control.

✅ Personal recommended range: CFG 2.5 → 6.5
(Experiment depending on prompt complexity + LoRAs)

If using IP-Adapter FaceID Plus (my V2 workflow)

• Keep CFG around ~1.5 for best results (FaceID tends to behave better at low CFG)

• If using ADetailer + IP-Adapter together:
  In ADetailer settings, set ControlNet model to “Passthrough”
  (so ADetailer doesn’t fight the identity guidance)

🚫 Suggested negative prompts (starter pack)

Use this as a baseline and adjust depending on your workflow:

Negatives:
((female, women, girl, pussy, breasts, vagina)), anime, toon \(style\), 2d, 3d, wide shot, long shot, distant subjects, painting, drawing, illustration, glitch, deformed, mutated, cross-eyed, ugly, disfigured

🧍 If the model keeps drifting female (quick fix)

Even with a male-biased checkpoint, some prompt setups can still pull female by default — especially things like:

• “couple portrait” without specifying genders clearly

• “romance scene” + soft aesthetic keywords

• certain style LoRAs that bias feminine faces/body language

• prompts that mention “beautiful / pretty” without adding male descriptors

Try adding this tag:

• (1boy:1) with weights

  • Start without weights first (1boy)

  • If no improvement, start at 1.0 and slowly increase

  • Usually works well up to around 2.5 (don’t overdo it or it can get weird)

📸 Sample image transparency (V2)

• Same base generation settings as Version 1

• V2 samples include an ADetailer pass (mainly for faces/details)

Version 1

What it’s good at

• Male-focused realism (better “male by default” behavior than my original bases)

• Gay couples / MM prompts

• Fantasy/escape vibes with realistic lighting and mood

• Works well for interspecies concepts when prompted (merman, etc.)

Merge credits (please support the original creators)

Base (merge):

• photorealisticAllPurpose_v40

• xlSatIORImitationOf_xlSATIorv72

“Spices” added:

• sdxlYamersRealistic5_v5Rundiffusion — (dreamy realism / mood)

• pornworksBadBoysPhoto_v06 — (male focus / MM bias)

• waiIllustriousSDXL_v150 — (slight prompt-following boost)

Links (kept together for easy copy/paste):

photorealisticAllPurpose_v40
https://civarchive.com/models/659428?modelVersionId=2113674

xlSatIORImitationOf_xlSATIorv72
https://civarchive.com/models/1713256?modelVersionId=2379373

sdxlYamersRealistic5_v5Rundiffusion
https://civarchive.com/models/127923/sdxl-yamers-realistic-5?modelVersionId=299716

pornworksBadBoysPhoto_v06
https://civarchive.com/models/1197994/pornworks-bad-boys-photo-realistic-gay-and-male-nsfw-aesthetics-sdxl-checkpoint

waiIllustriousSDXL_v150
https://civarchive.com/models/827184?modelVersionId=2167369

Recommended settings (what my samples use)

WebUI: ForgeNeo

• Sampler: DPM++ 2M SDE (recommended) / DPM++ 3M SDE

• Schedule: Karras

• Steps: 40–50

• Resolution: 896 × 1344

All sample images were generated with the above settings.
No ADetailer, ControlNet, or other helper tools were used.

More Examples (My Galleries)

You can see more images I’ve created (including male-focused + interspecies couples) on my pages below:

DeviantArt: https://www.deviantart.com/moodpixel

Patreon: http://patreon.com/MoodPixel

Note: Not every image on these pages is a “raw checkpoint-only” result. Some posts include additional workflows (for example: LoRAs, inpainting, ADetailer, ControlNet, upscalers, or post-processing). The sample images attached to this model page are the closest representation of the checkpoint using only the settings listed above.

Notes / Known quirks

• This is my first public merge, so if you hit odd edge cases, consider it “merge-life” and feel free to experiment.

• After heavier merges, some style LoRAs can feel weaker. If that happens, try:

  • increasing LoRA strength slightly

  • lowering CFG a bit

  • using img2img with low denoise for style passes

Important content note

One ingredient used for stronger male bias is NSFW-leaning, so this checkpoint may sometimes “suggest” more skin than requested depending on prompts. If you want clean outputs, use clear clothing terms and negatives like:
nude, naked, nsfw, explicit

I recommend ADetailer for the faces and other details.