From 2937fdd823dd103576c65584ab0b61e1244bdad3 Mon Sep 17 00:00:00 2001
From: Concedo <39025047+LostRuins@users.noreply.github.com>
Date: Sun, 24 May 2026 18:34:24 +0800
Subject: [PATCH] make send as reference image default for img2img, but with
smart fallback for older models.
---
embd_res/kcpp_sdui.embd | 32 +++----
otherarch/sdcpp/kcpp_sd_extensions.h | 1 +
otherarch/sdcpp/sdtype_adapter.cpp | 119 ++++++++++++++++-----------
otherarch/sdcpp/stable-diffusion.cpp | 1 +
4 files changed, 87 insertions(+), 66 deletions(-)
diff --git a/embd_res/kcpp_sdui.embd b/embd_res/kcpp_sdui.embd
index 56926be44..5bb2deb33 100644
--- a/embd_res/kcpp_sdui.embd
+++ b/embd_res/kcpp_sdui.embd
@@ -5,18 +5,18 @@
Stable UI for KoboldCpp
-
diff --git a/otherarch/sdcpp/kcpp_sd_extensions.h b/otherarch/sdcpp/kcpp_sd_extensions.h
index 5369d788c..cdf3d194a 100644
--- a/otherarch/sdcpp/kcpp_sd_extensions.h
+++ b/otherarch/sdcpp/kcpp_sd_extensions.h
@@ -15,6 +15,7 @@ namespace kcpp_sd {
bool is_sd1;
bool is_sd2;
bool is_sdxs;
+ bool is_sdxl;
bool is_wan;
bool is_zimage;
int vae_scale_factor;
diff --git a/otherarch/sdcpp/sdtype_adapter.cpp b/otherarch/sdcpp/sdtype_adapter.cpp
index fd5787f7e..0d3ca91c7 100644
--- a/otherarch/sdcpp/sdtype_adapter.cpp
+++ b/otherarch/sdcpp/sdtype_adapter.cpp
@@ -974,6 +974,12 @@ static std::string raw_image_to_png_base64(const sd_image_t& img, std::string pa
return result;
}
+bool supports_reference_images(kcpp_sd::model_info info)
+{
+ bool supported = (info.is_wan || info.is_qwenimg || info.is_flux2 || info.is_kontext || photomaker_enabled);
+ return supported;
+}
+
sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
{
if(sd_ctx == nullptr || sd_params == nullptr)
@@ -1055,6 +1061,16 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
}
}
+ //if a single extra image is provided, mask is NOT provided, and img2img image is NOT provided
+ //and it's a (SD1.5, SDXL) model that doesn't support extra images (see extra_image_data later)
+ //swap extra image data into img2img instead (graceful fallback)
+ if(!supports_reference_images(info) && extra_image_data.size()==1 && !is_img2img && img2img_mask=="")
+ {
+ is_img2img = true;
+ img2img_data = extra_image_data[0];
+ extra_image_data.clear();
+ }
+
if(info.is_wan && extra_image_data.size()==0 && is_img2img)
{
extra_image_data.push_back(img2img_data);
@@ -1135,42 +1151,11 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
{
int nx2, ny2, nc2;
int desiredchannels = 3;
- if(info.is_wan)
+ if(supports_reference_images(info))
{
- uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2,img2imgW,img2imgH,3);
- if(loaded)
+ if(info.is_wan)
{
- input_extraimage_buffers.push_back(loaded);
- sd_image_t extraimage_reference;
- extraimage_reference.width = nx2;
- extraimage_reference.height = ny2;
- extraimage_reference.channel = desiredchannels;
- extraimage_reference.data = loaded;
- wan_imgs.push_back(extraimage_reference);
- }
- }
- else if(info.is_qwenimg || info.is_flux2)
- {
- uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2);
- if(loaded)
- {
- //kcpp fix: qwen image can stack overflow and crash when ref images exceed
- // a total res of 512x512 = 262144, so we downscale if that's the case
- // kcpp edit 2mar2026: this seems to be better now, so limit to 1024x1024 instead
- int tgtx = nx2;
- int tgty = ny2;
- int res_lim_crash = 1024 * 1024;
- if (nx2 * ny2 > res_lim_crash)
- {
- float factor = sqrtf((float)res_lim_crash / ((float)nx2 * (float)ny2));
- tgtx = (int)(nx2 * factor);
- tgty = (int)(ny2 * factor);
- if (!sd_is_quiet && sddebugmode == 1)
- {
- printf("\nResized RefImg %dx%d to %dx%d", nx2, ny2, tgtx, tgty);
- }
- loaded = resize_image(loaded, nx2, ny2, tgtx, tgty);
- }
+ uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2,img2imgW,img2imgH,3);
if(loaded)
{
input_extraimage_buffers.push_back(loaded);
@@ -1179,28 +1164,62 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
extraimage_reference.height = ny2;
extraimage_reference.channel = desiredchannels;
extraimage_reference.data = loaded;
- reference_imgs.push_back(extraimage_reference);
+ wan_imgs.push_back(extraimage_reference);
}
}
- }
- else if (info.is_kontext || photomaker_enabled)
- {
- uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2);
- if(loaded)
+ else if(info.is_qwenimg || info.is_flux2)
{
- input_extraimage_buffers.push_back(loaded);
- sd_image_t extraimage_reference;
- extraimage_reference.width = nx2;
- extraimage_reference.height = ny2;
- extraimage_reference.channel = desiredchannels;
- extraimage_reference.data = loaded;
- if(info.is_kontext)
+ uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2);
+ if(loaded)
{
- reference_imgs.push_back(extraimage_reference);
+ //kcpp fix: qwen image can stack overflow and crash when ref images exceed
+ // a total res of 512x512 = 262144, so we downscale if that's the case
+ // kcpp edit 2mar2026: this seems to be better now, so limit to 1024x1024 instead
+ int tgtx = nx2;
+ int tgty = ny2;
+ int res_lim_crash = 1024 * 1024;
+ if (nx2 * ny2 > res_lim_crash)
+ {
+ float factor = sqrtf((float)res_lim_crash / ((float)nx2 * (float)ny2));
+ tgtx = (int)(nx2 * factor);
+ tgty = (int)(ny2 * factor);
+ if (!sd_is_quiet && sddebugmode == 1)
+ {
+ printf("\nResized RefImg %dx%d to %dx%d", nx2, ny2, tgtx, tgty);
+ }
+ loaded = resize_image(loaded, nx2, ny2, tgtx, tgty);
+ }
+ if(loaded)
+ {
+ input_extraimage_buffers.push_back(loaded);
+ sd_image_t extraimage_reference;
+ extraimage_reference.width = nx2;
+ extraimage_reference.height = ny2;
+ extraimage_reference.channel = desiredchannels;
+ extraimage_reference.data = loaded;
+ reference_imgs.push_back(extraimage_reference);
+ }
}
- else
+ }
+ else if (info.is_kontext || photomaker_enabled)
+ {
+ uint8_t * loaded = load_image_from_b64(extra_image_data[i],nx2,ny2);
+ if(loaded)
{
- photomaker_imgs.push_back(extraimage_reference);
+ input_extraimage_buffers.push_back(loaded);
+ sd_image_t extraimage_reference;
+ extraimage_reference.width = nx2;
+ extraimage_reference.height = ny2;
+ extraimage_reference.channel = desiredchannels;
+ extraimage_reference.data = loaded;
+ if(info.is_kontext)
+ {
+ reference_imgs.push_back(extraimage_reference);
+ }
+ else
+ {
+ photomaker_imgs.push_back(extraimage_reference);
+ }
}
}
}
diff --git a/otherarch/sdcpp/stable-diffusion.cpp b/otherarch/sdcpp/stable-diffusion.cpp
index fe4bcfff4..68b40c1b1 100644
--- a/otherarch/sdcpp/stable-diffusion.cpp
+++ b/otherarch/sdcpp/stable-diffusion.cpp
@@ -5637,6 +5637,7 @@ namespace kcpp_sd {
res.is_sdxs = (loadedsdver == SDVersion::VERSION_SDXS_512_DS || loadedsdver == SDVersion::VERSION_SDXS_09);
res.is_sd1 = (loadedsdver == SDVersion::VERSION_SD1);
res.is_sd2 = (loadedsdver == SDVersion::VERSION_SD2);
+ res.is_sdxl = sd_version_is_sdxl((SDVersion)loadedsdver);
res.vae_scale_factor = ctx->sd->get_vae_scale_factor();
res.spatial_multiple = get_spatial_multiple(ctx);
return res;