@@ -679,17 +679,32 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
679679 }
680680
681681 std::vector<sd_image_t > reference_imgs;
682+ std::vector<sd_image_t > wan_imgs;
682683 bool is_wan = (loadedsdver == SDVersion::VERSION_WAN2 || loadedsdver == SDVersion::VERSION_WAN2_2_I2V || loadedsdver == SDVersion::VERSION_WAN2_2_TI2V);
683684 bool is_kontext = (loadedsdver==SDVersion::VERSION_FLUX && !loaded_model_is_chroma (sd_ctx));
684- if (extra_image_data.size ()>0 && (is_wan || is_kontext) )
685+ if (extra_image_data.size ()>0 )
685686 {
686- for ( int i= 0 ;i<extra_image_data. size ();++i )
687+ if (is_kontext )
687688 {
688- reference_imgs.push_back (extraimage_references[i]);
689+ for (int i=0 ;i<extra_image_data.size ();++i)
690+ {
691+ reference_imgs.push_back (extraimage_references[i]);
692+ }
693+ if (!sd_is_quiet && sddebugmode==1 )
694+ {
695+ printf (" \n Image Gen: Using %d reference images\n " ,reference_imgs.size ());
696+ }
689697 }
690- if (!sd_is_quiet && sddebugmode== 1 )
698+ if (is_wan )
691699 {
692- printf (" \n Image Gen: Using %d reference images\n " ,reference_imgs.size ());
700+ for (int i=0 ;i<extra_image_data.size ();++i)
701+ {
702+ wan_imgs.push_back (extraimage_references[i]);
703+ }
704+ if (!sd_is_quiet && sddebugmode==1 )
705+ {
706+ printf (" \n Image Gen: Using %d video reference images\n " ,wan_imgs.size ());
707+ }
693708 }
694709 }
695710
@@ -727,7 +742,6 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
727742
728743 params.ref_images = reference_imgs.data ();
729744 params.ref_images_count = reference_imgs.size ();
730-
731745 params.pm_params .id_images = photomaker_imgs.data ();
732746 params.pm_params .id_images_count = photomaker_imgs.size ();
733747
@@ -752,15 +766,15 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
752766 vid_gen_params.strength = params.strength ;
753767 vid_gen_params.seed = params.seed ;
754768 vid_gen_params.video_frames = vid_req_frames;
755- if (reference_imgs .size ()>0 )
769+ if (wan_imgs .size ()>0 )
756770 {
757- if (reference_imgs .size ()>=1 )
771+ if (wan_imgs .size ()>=1 )
758772 {
759- vid_gen_params.init_image = reference_imgs [0 ];
773+ vid_gen_params.init_image = wan_imgs [0 ];
760774 }
761- if (reference_imgs .size ()>=2 )
775+ if (wan_imgs .size ()>=2 )
762776 {
763- vid_gen_params.end_image = reference_imgs [1 ];
777+ vid_gen_params.end_image = wan_imgs [1 ];
764778 }
765779 }
766780 if (!sd_is_quiet && sddebugmode==1 )
@@ -775,7 +789,7 @@ sd_generation_outputs sdtype_generate(const sd_generation_inputs inputs)
775789 << " \n STRENGTH:" << vid_gen_params.strength
776790 << " \n FRAMES:" << vid_gen_params.video_frames
777791 << " \n CTRL_FRM:" << vid_gen_params.control_frames_size
778- << " \n REF_IMGS :" << reference_imgs .size ()
792+ << " \n INIT_IMGS :" << wan_imgs .size ()
779793 << " \n\n " ;
780794 printf (" %s" , ss.str ().c_str ());
781795 }
0 commit comments