A LoRA adapter for FLUX 2 Klein 9B fine-tuned for high-quality face swapping. Given a face source image and a body/scene image, the model produces a natural face swap preserving identity, lighting, and skin tone.

Prompt

Use the following prompt for best results:

face_swap: take the face from Picture 1 and swap it onto the person in Picture 2. the result should show the exact face and features from Picture 1 on the body, pose, and scene from Picture 2, with natural skin tone blending and matching lighting.

Picture 1: Face source image (the identity to transfer)
Picture 2: Body/scene image (the target pose and environment)

How to Use

Try it on fal

Open on fal.ai

Via fal.ai API

python

import fal_client

def on_queue_update(update):
    if isinstance(update, fal_client.InProgress):
        for log in update.logs:
           print(log["message"])

result = fal_client.subscribe(
    "fal-ai/flux-2-klein/9b/base/edit/lora",
    arguments={
        "prompt": "face_swap: take the face from Picture 1 and swap it onto the person in Picture 2. "
                  "the result should show the exact face and features from Picture 1 on the body, pose, "
                  "and scene from Picture 2, with natural skin tone blending and matching lighting.",
        "model_name": None,
        "loras": [{
            "path": "https://huggingface.co/ilkerzgi/face-swap/resolve/main/FS2_V2_r32_800_pytorch_lora_weights.safetensors",
            "scale": 1
        }],
        "embeddings": [],
        "image_urls": [
            "https://example.com/face_source.png",
            "https://example.com/body_scene.png"
        ],
        "num_images": 4
    },
    with_logs=True,
    on_queue_update=on_queue_update,
)
print(result)

Evaluation Overview

Test Cases	LoRA Variants	Image Models	Total Generations	Dataset Sizes
25	9	3	300	300 / 800

9 LoRA variants were trained with different hyperparameters and evaluated against 3 image models across 25 diverse test subjects.

Training Configurations

Variant	Samples	Rank	Steps	LR	Guidance	Avg Latency
V1 r16 (300)	300	16	3000	5e-5	1	6.0s
V2 r32 (300)	300	32	3500	5e-5	1	6.1s
V3 r32+cfg (300)	300	32	3500	5e-5	3.5	6.6s
V1 r16 (800)	800	16	3000	5e-5	1	6.4s
V2 r32 (800)	800	32	3500	5e-5	1	6.3s
V3 r32+cfg (800)	800	32	3500	5e-5	3.5	6.2s
V4 r32 long (800)	800	32	5000	5e-5	1	6.4s
V5 r64 (800)	800	64	4000	3e-5	1	6.4s
V6 r32 slow (800)	800	32	5000	1e-5	1	6.4s

Image Models

Model	Avg Latency
Nano Banana 2	28.8s
Nano Banana Pro	33.3s
GPT Image 1.5	30.3s

Latency Comparison

Model	Avg Latency
V1 r16 (300)	6.0s
V2 r32 (300)	6.1s
V3 r32+cfg (800)	6.2s
V2 r32 (800)	6.3s
V6 r32 slow (800)	6.4s
V4 r32 long (800)	6.4s
V5 r64 (800)	6.4s
V1 r16 (800)	6.4s
V3 r32+cfg (300)	6.6s
Nano Banana 2	28.8s
GPT Image 1.5	30.3s
Nano Banana Pro	33.3s

All LoRA variants average ~6-7s, while image models average ~29-33s.

Model Weights

Variant	Rank	Samples	Download
V1 r16 (300)	16	300	FS_V1_r16_300_pytorch_lora_weights.safetensors
V2 r32 (300)	32	300	FS_V2_r32_300_pytorch_lora_weights.safetensors
V3 r32+cfg (300)	32	300	FS_V3_r32g_300_pytorch_lora_weights.safetensors
V1 r16 (800)	16	800	FS2_V1_r16_800_pytorch_lora_weights.safetensors
V2 r32 (800)	32	800	FS2_V2_r32_800_pytorch_lora_weights.safetensors
V3 r32+cfg (800)	32	800	FS2_V3_r32g_800_pytorch_lora_weights.safetensors
V4 r32 long (800)	32	800	FS2_V4_r32L_800_pytorch_lora_weights.safetensors
V5 r64 (800)	64	800	FS2_V5_r64_800_pytorch_lora_weights.safetensors
V6 r32 slow (800)	32	800	FS2_V6_r32s_800_pytorch_lora_weights.safetensors

Test Cases (25)

#1 — middle-aged woman with short stylish grey hair and blue eyes, elegant bone structure

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	6.7s	8.9s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

7.3s	6.6s	6.6s	6.6s	6.6s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

35.3s	39.6s	27.7s

#2 — young woman with long straight brown hair parted in the middle and brown eyes

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.6s	6.1s	7.1s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	7.1s	6.8s	7.1s	9.4s	6.8s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

24.0s	32.5s	44.6s

#3 — older woman with short white curly hair and warm brown eyes

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.1s	6.3s	5.9s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.6s	6.6s	6.6s	6.6s	6.6s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.0s	30.4s	50.6s

#4 — young African American woman with braided hair and warm brown eyes

#4 — young African American woman with braided hair and warm brown eyes

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

7.5s	7.4s	5.8s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.6s	6.6s	7.6s	6.6s	5.7s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

31.0s	36.6s	28.0s

#5 — young woman with short blonde bob haircut and bright blue eyes, Scandinavian features

#5 — young woman with short blonde bob haircut and bright blue eyes, Scandinavian features

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

8.0s	6.6s	6.1s	6.7s	6.9s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

26.5s	41.0s	25.3s

#6 — young Hispanic man with wavy dark hair and brown eyes, trimmed goatee

#6 — young Hispanic man with wavy dark hair and brown eyes, trimmed goatee

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.1s	7.1s	6.1s	6.6s	6.6s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.4s	42.3s	24.0s

#7 — young South Asian man with dark thick hair and brown eyes, short neat beard

#7 — young South Asian man with dark thick hair and brown eyes, short neat beard

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	6.6s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.6s	6.6s	6.6s	7.1s	7.5s	7.1s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

23.0s	31.3s	25.4s

#8 — young man with a buzz cut and hazel eyes, tanned skin

#8 — young man with a buzz cut and hazel eyes, tanned skin

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.2s	6.6s	6.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.1s	6.6s	6.6s	6.8s	5.5s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

28.5s	39.8s	26.6s

#9 — young Middle Eastern man with dark wavy hair and olive skin, prominent eyebrows

#9 — young Middle Eastern man with dark wavy hair and olive skin, prominent eyebrows

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.9s	6.6s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.1s	7.3s	8.9s	6.6s	5.7s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.3s	44.6s	66.6s

#10 — young man with curly black hair and deep brown eyes, clean-shaven

#10 — young man with curly black hair and deep brown eyes, clean-shaven

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.4s	5.7s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	7.1s	6.1s	7.0s	6.7s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

43.9s	31.3s	25.2s

#11 — young African American woman with short natural TWA hairstyle and dark brown eyes, high cheekbones

#11 — young African American woman with short natural TWA hairstyle and dark brown eyes, high cheekbones

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	6.7s	6.6s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

9.8s	7.5s	5.7s	6.6s	6.1s	7.5s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

26.1s	23.3s	25.3s

#12 — young woman with tight blonde curls, dark brown skin, bright smile, dimples

#12 — young woman with tight blonde curls, dark brown skin, bright smile, dimples

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.4s	10.3s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.7s	6.1s	6.6s	6.2s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

30.6s	25.8s	25.4s

#13 — young woman with tight blonde curls, dark brown skin, bright smile, dimples

#13 — young woman with tight blonde curls, dark brown skin, bright smile, dimples

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

7.1s	5.7s	7.0s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	6.6s	5.7s	7.0s	5.8s	6.8s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

24.1s	24.4s	26.6s

#14 — middle-aged woman with dark skin, close-cropped natural hair, elegant gold jewelry

#14 — middle-aged woman with dark skin, close-cropped natural hair, elegant gold jewelry

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	7.1s	7.5s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

7.1s	5.7s	7.1s	5.7s	7.1s	7.1s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

19.4s	24.7s	28.1s

#15 — young Nordic man with platinum blonde hair, ice blue eyes, sharp cheekbones, clean shaven

#15 — young Nordic man with platinum blonde hair, ice blue eyes, sharp cheekbones, clean shaven

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	6.6s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.6s	5.7s	5.7s	5.7s	7.5s	7.1s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

26.4s	42.2s	25.3s

#16 — middle-aged Hispanic woman with dark wavy hair streaked with grey, warm brown eyes, round face

#16 — middle-aged Hispanic woman with dark wavy hair streaked with grey, warm brown eyes, round face

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	10.3s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.4s	5.7s	5.7s	7.5s	7.1s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

45.2s	34.2s	24.0s

#17 — young Pacific Islander woman with long wavy dark hair, brown eyes, strong features

#17 — young Pacific Islander woman with long wavy dark hair, brown eyes, strong features

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	7.1s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	8.1s	5.7s	5.7s	5.7s	6.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

20.0s	28.0s	27.3s

#18 — older Eastern European man with a thick grey mustache, lined face, piercing grey eyes

#18 — older Eastern European man with a thick grey mustache, lined face, piercing grey eyes

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.9s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.7s	5.7s	5.7s	5.7s	5.2s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

20.7s	24.0s	26.7s

#19 — older African American woman with silver locs, deep brown eyes, and laugh lines, regal posture

#19 — older African American woman with silver locs, deep brown eyes, and laugh lines, regal posture

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.2s	6.1s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.9s	5.7s	5.7s	5.7s	6.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.7s	58.2s	24.8s

#20 — older Chinese woman with jet black dyed hair, round face, gentle brown eyes, pearl earrings

#20 — older Chinese woman with jet black dyed hair, round face, gentle brown eyes, pearl earrings

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	6.6s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.8s	5.7s	5.7s	7.1s	5.6s	5.6s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

35.0s	27.9s	28.3s

#21 — middle-aged woman with fiery red curly hair and freckled pale skin, green eyes, strong jawline

#21 — middle-aged woman with fiery red curly hair and freckled pale skin, green eyes, strong jawline

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.6s	5.7s	5.6s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.7s	5.7s	5.7s	6.6s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

41.1s	36.6s	27.9s

#22 — young Southeast Asian man with black hair, smooth brown skin, sharp jawline

#22 — young Southeast Asian man with black hair, smooth brown skin, sharp jawline

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.7s	5.7s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.7s	6.6s	5.7s	5.7s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.3s	28.8s	25.3s

#23 — middle-aged woman with natural grey afro, dark skin, reading glasses on nose

#23 — middle-aged woman with natural grey afro, dark skin, reading glasses on nose

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.6s	5.7s	5.7s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

6.1s	5.7s	5.8s	5.6s	6.6s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

34.1s	22.0s	24.6s

#24 — middle-aged Hispanic woman with dark wavy hair streaked with grey, warm brown eyes, round face

#24 — middle-aged Hispanic woman with dark wavy hair streaked with grey, warm brown eyes, round face

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

5.6s	5.7s	6.1s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

7.8s	5.7s	5.9s	5.7s	5.7s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

31.9s	24.7s	26.0s

#25 — older East Asian man with thinning grey hair, wire-rimmed glasses, kind wrinkled eyes

#25 — older East Asian man with thinning grey hair, wire-rimmed glasses, kind wrinkled eyes

Face Source	Body / Scene	Ground Truth

LoRA — 300 Samples

V1 r16 (300)	V2 r32 (300)	V3 r32+cfg (300)

6.6s	5.7s	6.0s

LoRA — 800 Samples

V1 r16 (800)	V2 r32 (800)	V3 r32+cfg (800)	V4 r32 long (800)	V5 r64 (800)	V6 r32 slow (800)

5.7s	5.7s	5.4s	5.7s	5.7s	5.7s

Image Models

Nano Banana 2	Nano Banana Pro	GPT Image 1.5

25.3s	37.3s	46.7s

Training Details

Base Model: FLUX 2 Klein 9B
Method: LoRA (Low-Rank Adaptation)
Trainer: fal-ai/flux-2-klein-9b-base-trainer
Datasets: Two custom face-swap paired datasets (300 and 800 samples)
Rank variants tested: 16, 32, 64
Training steps: 3,000 – 5,000
Learning rates: 1e-5 to 5e-5
Guidance scale: 1.0 and 3.5

Author

Ilker Izgi — Creative Engineer at fal