usamaehsan/swap-sd 🖼️🔢📝✓ → 🖼️

▶️ 6.6K runs 📅 Feb 2024 ⚙️ Cog 0.9.4
image-face-swap image-to-image

About

Experimental & for non-commercial use only

Example Output

Prompt:

"This is Eric Draven The Crow Movie 1994 version, Brandon Lee, realistic, 8k, the crow bird, portrait"

Output

Example output

Performance Metrics

14.58s Prediction Time
153.10s Total Time
All Input Parameters 
{
  "image": "https://replicate.delivery/pbxt/KOVEvXzae3nKcjxgRTsHrfdWN88WaLuPquXrognA336KLDuL/53894527%20%281%29.jpg",
  "prompt": "This is Eric Draven The Crow Movie 1994 version, Brandon Lee, realistic, 8k, the crow bird, portrait",
  "max_width": 512,
  "max_height": 512,
  "guidance_scale": 0.1,
  "negative_prompt": "Longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality",
  "ip_adapter_scale": 0.8,
  "num_inference_steps": 4,
  "disable_safety_check": false
}
Input Parameters
image Type: string
Input image
width Type: integerDefault: 512
Max width/Resolution of image
height Type: integerDefault: 512
Max height/Resolution of image
prompt (required) Type: string
Prompt - using compel, use +++ to increase words weight:: doc: https://github.com/damian0815/compel/tree/main/doc || https://invoke-ai.github.io/InvokeAI/features/PROMPTS/#attention-weighting
swap_face Type: booleanDefault: true
pose_image Type: string
pose image
pose_scale Type: numberDefault: 0.8Range: 0 - 5
Scale for open pose controlnet
use_gfpgan Type: booleanDefault: true
gfpgan to enhance face
guidance_scale Type: numberDefault: 2Range: 0.1 - 30
Scale for classifier-free guidance
negative_prompt Type: stringDefault: Longbody, lowres, bad anatomy, bad hands, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality
Negative prompt - using compel, use +++ to increase words weight//// negative-embeddings available ///// FastNegativeV2 , boring_e621_v4 , verybadimagenegative_v1 || to use them, write their keyword in negative prompt
ip_adapter_scale Type: numberDefault: 0.8Range: 0 - 5
Scale for IP adapter
num_inference_steps Type: integerDefault: 20
Steps to run denoising
disable_safety_check Type: booleanDefault: false
Disable safety check. Use at your own risk!
use_pose_image_resolution Type: booleanDefault: true
image will be generated in pose' width and height
Output Schema

Output

Type: array • Items Type: string • Items Format: uri

Example Execution Logs 
0%|          | 0/4 [00:00<?, ?it/s]
 25%|██▌       | 1/4 [00:00<00:00,  4.75it/s]
100%|██████████| 4/4 [00:00<00:00, 14.18it/s]
100%|██████████| 4/4 [00:00<00:00, 12.34it/s]
/root/.pyenv/versions/3.11.8/lib/python3.11/site-packages/insightface/utils/transform.py:68: FutureWarning: `rcond` parameter will change to the default of machine precision times ``max(M, N)`` where M and N are the input matrix dimensions.
To use the future default and silence this warning we advise to pass `rcond=None`, to keep using the old, explicitly pass `rcond=-1`.
P = np.linalg.lstsq(X_homo, Y)[0].T # Affine matrix. 3 x 4
Version Details
Version ID
ce62b1e250fd5be8614cb3437e6f61e4f7dd56be2cb7c959d890239b9f7b1294
Version Created
March 3, 2024
Run on Replicate →