add character reference

app.py

@@ -185,6 +185,10 @@ def control_ui():
                     coords_y[i].input(fn=change_one, inputs=[char_coords_y, coords_y[i]], outputs=char_coords_y)
             num_chars.change(lambda: gr.Accordion(open=True), inputs=None, outputs=chars)
             auto_pos.change(lambda: gr.Accordion(open=True), inputs=None, outputs=chars)
+        with gr.Accordion('角色参考（仅nai4.5有效）', open=False, elem_id="v4.5char_tab") as v45char_tab:
+            chr_image = gr.Image(label="上传图片", value=None, sources=["upload", "clipboard", "webcam"], interactive=True, type="pil", show_share_button=False)
+            fidelity = gr.Slider(label='参考度', value=1, minimum=0, maximum=1, step=0.05)
+            style_aware = gr.Checkbox(label='参考风格', value=True)
         with gr.Accordion('高级选项', open=False):
             scheduler = gr.Dropdown(
                 choices=[
@@ -212,10 +216,10 @@ def control_ui():
     return gen_btn, stop_btn, \
         [model, prompt, quality_tags, neg_prompt, seed, scale, width, height, steps, sampler, scheduler, smea, dyn, dyn_threshold, cfg_rescale, variety, \
         ref_images, info_extracts, ref_strs, vibe_files, str_norm, i2i_image, i2i_str, i2i_noise, overlay, inp_img, selection, \
-        num_chars, auto_pos, char_prompts, char_ucs, char_coords_x, char_coords_y, legacy], \
+        num_chars, auto_pos, char_prompts, char_ucs, char_coords_x, char_coords_y, legacy, chr_image, fidelity, style_aware], \
         [save, rand_seed, reuse_seed, reuse_img_i2i, reuse_img_inp, vibe_tab, i2i_tab, rerender]
 
-def generate(model, prompt, quality_tags, neg_prompt, seed, scale, width, height, steps, sampler, scheduler, smea, dyn, dyn_threshold, cfg_rescale, variety, ref_images, info_extracts, ref_strs, vibe_files, str_norm, i2i_image, i2i_str, i2i_noise, overlay, inp_img, selection, num_chars, auto_pos, char_prompts, char_ucs, char_coords_x, char_coords_y, legacy):
+def generate(model, prompt, quality_tags, neg_prompt, seed, scale, width, height, steps, sampler, scheduler, smea, dyn, dyn_threshold, cfg_rescale, variety, ref_images, info_extracts, ref_strs, vibe_files, str_norm, i2i_image, i2i_str, i2i_noise, overlay, inp_img, selection, num_chars, auto_pos, char_prompts, char_ucs, char_coords_x, char_coords_y, legacy, chr_image, fidelity, style_aware):
     global today_count
     set_token(os.environ.get('token'))
     img_data, payload = generate_novelai_image(
@@ -223,7 +227,8 @@ def generate(model, prompt, quality_tags, neg_prompt, seed, scale, width, height
         width, height, steps, sampler, scheduler, 
         smea, dyn, dyn_threshold, cfg_rescale, variety, ref_images, info_extracts, ref_strs, vibe_files, str_norm,
         i2i_image, i2i_str, i2i_noise, overlay, inp_img, selection,
-        auto_pos, char_prompts[:num_chars], char_ucs[:num_chars], char_coords_x[:num_chars], char_coords_y[:num_chars], legacy
+        auto_pos, char_prompts[:num_chars], char_ucs[:num_chars], char_coords_x[:num_chars], char_coords_y[:num_chars], legacy,
+        chr_image, fidelity, style_aware
     )
     if not isinstance(img_data, bytes):
         return gr.Image(value=None), payload
@@ -281,7 +286,7 @@ def display_vibe_thumbs(vibe_files):
             thumbs.append(base642image(data))
     return gr.Gallery(value=thumbs, visible=True)
 
-def update_btn_cost(model, w, h, s, sm, dyn, ref_imgs, i2i_img, i2i_str, inp_img, selection):
+def update_btn_cost(model, w, h, s, sm, dyn, ref_imgs, i2i_img, i2i_str, inp_img, selection, chr_img):
     if selection == 'i2i' and i2i_img != None:
         cost = calculate_cost(w, h, s, False, False, i2i_str)
     elif selection == 'inp' and inp_img['background'] != None:
@@ -290,6 +295,8 @@ def update_btn_cost(model, w, h, s, sm, dyn, ref_imgs, i2i_img, i2i_str, inp_img
         cost = calculate_cost(w, h, s, sm, dyn)
     if len(ref_imgs) and model.startswith('nai-diffusion-4'):
         cost += len(ref_imgs) * 2
+    if chr_img != None and model.startswith('nai-diffusion-4-5'):
+        cost += 5
     return gr.Button(value=f"生成(预计消耗{cost}点数)")
 
 def main_ui():
@@ -299,7 +306,7 @@ def main_ui():
                 gen_btn, stop_btn, paras, others = control_ui()
             with gr.Column():
                 image, info, send_dtool = preview_ui()
-    cost_list = [paras[0]] + paras[6:9] + paras[11:13] + [paras[17]] + paras[21:23] + paras[25:27]
+    cost_list = [paras[0]] + paras[6:9] + paras[11:13] + [paras[17]] + paras[21:23] + paras[25:27] + [paras[34]]
     for component in cost_list:
         component.change(update_btn_cost, inputs=cost_list, outputs=gen_btn)
     gen = gen_btn.click(lambda: (gr.Button(visible=False), gr.Button(visible=True)), inputs=None, outputs=[gen_btn, stop_btn]).then(

utils.py

@@ -77,7 +77,10 @@ def generate_novelai_image(
     char_ucs=[],
     char_coords_x=[],
     char_coords_y=[],
-    legacy=False
+    legacy=False,
+    chr_image=None,
+    fidelity=1,
+    style_aware=True
 ):
     # Assign a random seed if seed is -1
     if seed == -1:
@@ -151,6 +154,10 @@ def generate_novelai_image(
         payload['parameters']['strength'] = i2i_str
         payload['parameters']['extra_noise_seed'] = seed
         payload["parameters"]['noise'] = i2i_noise
+    if chr_image != None:
+        payload['parameters']['director_reference_images'] = [image2base64(chr_image)]
+        payload['parameters']['director_reference_descriptions'] = [{'caption': {'base_caption': 'character&style' if style_aware else 'character', 'char_captions': []}, 'legacy_uc': False}]
+        payload['parameters']['director_reference_secondary_strength_values'] = [1 - fidelity]
     # Send the POST request
     try:
         response = requests.post(url, json=payload, headers=headers, timeout=180)