diff --git a/expose.h b/expose.h
index fb25fc8ca..149117987 100644
--- a/expose.h
+++ b/expose.h
@@ -168,6 +168,7 @@ struct sd_generation_inputs
const char * prompt = nullptr;
const char * negative_prompt = nullptr;
const char * init_images = "";
+ const char * mask = "";
const float denoising_strength = 0.0f;
const float cfg_scale = 0.0f;
const int sample_steps = 0;
diff --git a/kcpp_sdui.embd b/kcpp_sdui.embd
index c9c83f638..f1f4a5efc 100644
--- a/kcpp_sdui.embd
+++ b/kcpp_sdui.embd
@@ -5,18 +5,18 @@
Stable UI for KoboldCpp
-
-
diff --git a/koboldcpp.py b/koboldcpp.py
index fe5463523..dc4dbce6e 100644
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -263,6 +263,7 @@ class sd_generation_inputs(ctypes.Structure):
_fields_ = [("prompt", ctypes.c_char_p),
("negative_prompt", ctypes.c_char_p),
("init_images", ctypes.c_char_p),
+ ("mask", ctypes.c_char_p),
("denoising_strength", ctypes.c_float),
("cfg_scale", ctypes.c_float),
("sample_steps", ctypes.c_int),
@@ -1484,6 +1485,7 @@ def sd_generate(genparams):
prompt = forced_posprompt
init_images_arr = genparams.get("init_images", [])
init_images = ("" if (not init_images_arr or len(init_images_arr)==0 or not init_images_arr[0]) else init_images_arr[0])
+ mask = genparams.get("mask", "")
denoising_strength = tryparsefloat(genparams.get("denoising_strength", 0.6))
cfg_scale = tryparsefloat(genparams.get("cfg_scale", 5))
sample_steps = tryparseint(genparams.get("steps", 20))
@@ -1520,6 +1522,7 @@ def sd_generate(genparams):
inputs.prompt = prompt.encode("UTF-8")
inputs.negative_prompt = negative_prompt.encode("UTF-8")
inputs.init_images = init_images.encode("UTF-8")
+ inputs.mask = "".encode("UTF-8") if not mask else mask.encode("UTF-8")
inputs.cfg_scale = cfg_scale
inputs.denoising_strength = denoising_strength
inputs.sample_steps = sample_steps
diff --git a/otherarch/sdcpp/sdtype_adapter.cpp b/otherarch/sdcpp/sdtype_adapter.cpp
index ff06bbe32..8762f8964 100644
--- a/otherarch/sdcpp/sdtype_adapter.cpp
+++ b/otherarch/sdcpp/sdtype_adapter.cpp
@@ -160,6 +160,7 @@ bool sdtype_load_model(const sd_load_model_inputs inputs) {
{
printf("With Custom Clip-G Model: %s\n",clipg_filename.c_str());
}
+ printf("\n");
//duplicated from expose.cpp
int cl_parseinfo = inputs.clblast_info; //first digit is whether configured, second is platform, third is devices
@@ -320,7 +321,7 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
std::string cleanprompt = clean_input_prompt(inputs.prompt);
std::string cleannegprompt = clean_input_prompt(inputs.negative_prompt);
std::string img2img_data = std::string(inputs.init_images);
- std::string img2img_mask = "";
+ std::string img2img_mask = std::string(inputs.mask);
std::string sampler = inputs.sample_method;
sd_params->prompt = cleanprompt;
@@ -506,9 +507,9 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
if(img2img_mask!="")
{
image_mask_buffer = kcpp_base64_decode(img2img_mask);
- input_mask_buffer = stbi_load_from_memory(image_mask_buffer.data(), image_mask_buffer.size(), &nx2, &ny2, &nc2, 3);
+ input_mask_buffer = stbi_load_from_memory(image_mask_buffer.data(), image_mask_buffer.size(), &nx2, &ny2, &nc2, 1);
// Resize the image
- int resok = stbir_resize_uint8(input_mask_buffer, nx, ny, 0, resized_mask_buf.data(), img2imgW, img2imgH, 0, img2imgC);
+ int resok = stbir_resize_uint8(input_mask_buffer, nx2, ny2, 0, resized_mask_buf.data(), img2imgW, img2imgH, 0, 1);
if (!resok) {
printf("\nKCPP SD: resize image failed!\n");
output.data = "";