Multi prompts feature

zoom.py

@@ -16,26 +16,25 @@ inpaint_model_list = [
 ]
 default_prompt = "A psychedelic jungle with trees that have glowing, fractal-like patterns, Simon stalenhag poster 1920s style, street level view, hyper futuristic, 8k resolution, hyper realistic"
 default_negative_prompt = "frames, borderline, text, charachter, duplicate, error, out of frame, watermark, low quality, ugly, deformed, blur"
-# TODO:
-# prompts = {
-#     0: "prompt1",
-#     7: "prompt2"
-# }
-
-custom_init_image = False
-init_image_address = "/init/image.jpeg"
 
 
 def zoom(
     model_id,
-    prompt,
+    prompts_array,
     negative_prompt,
     num_outpainting_steps,
     guidance_scale,
     num_inference_steps,
     custom_init_image
 ):
-
+    prompts = {}
+    for x in prompts_array:
+        try:
+            key = int(x[0])
+            value = str(x[1])
+            prompts[key] = value
+        except ValueError:
+            pass
     pipe = StableDiffusionInpaintPipeline.from_pretrained(
         model_id,
         torch_dtype=torch.float16,
@@ -57,18 +56,19 @@ def zoom(
     mask_image = np.array(current_image)[:, :, 3]
     mask_image = Image.fromarray(255-mask_image).convert("RGB")
     current_image = current_image.convert("RGB")
-    if(custom_init_image): 
-      current_image = custom_init_image.resize((width, height), resample=Image.LANCZOS)
+    if (custom_init_image):
+        current_image = custom_init_image.resize(
+            (width, height), resample=Image.LANCZOS)
     else:
-      init_images = pipe(prompt=prompt,  # TODO: prompt=prompts[max(k for k in prompts.keys() if k >= 0)],
-                        negative_prompt=negative_prompt,
-                        image=current_image,
-                        guidance_scale=guidance_scale,
-                        height=height,
-                        width=width,
-                        mask_image=mask_image,
-                        num_inference_steps=num_inference_steps)[0]
-      current_image = init_images[0]
+        init_images = pipe(prompt=prompts[min(k for k in prompts.keys() if k >= 0)],
+                           negative_prompt=negative_prompt,
+                           image=current_image,
+                           guidance_scale=guidance_scale,
+                           height=height,
+                           width=width,
+                           mask_image=mask_image,
+                           num_inference_steps=num_inference_steps)[0]
+        current_image = init_images[0]
     mask_width = 128
     num_interpol_frames = 30
 
@@ -91,7 +91,7 @@ def zoom(
 
         # inpainting step
         current_image = current_image.convert("RGB")
-        images = pipe(prompt=prompt,  # TODO: prompt=prompts[max(k for k in prompts.keys() if k <= i)],
+        images = pipe(prompt=prompts[max(k for k in prompts.keys() if k <= i)],
                       negative_prompt=negative_prompt,
                       image=current_image,
                       guidance_scale=guidance_scale,
@@ -125,8 +125,8 @@ def zoom(
             interpol_image.paste(prev_image_fix_crop, mask=prev_image_fix_crop)
 
             all_frames.append(interpol_image)
-
         all_frames.append(current_image)
+        # interpol_image.show()
     video_file_name = "infinite_zoom_" + str(time.time())
     fps = 30
     save_path = video_file_name + ".mp4"
@@ -137,15 +137,20 @@ def zoom(
                 start_frame_dupe_amount, last_frame_dupe_amount)
     return save_path
 
+
 def zoom_app():
     with gr.Blocks():
         with gr.Row():
             with gr.Column():
 
-                outpaint_prompt = gr.Textbox(
-                    lines=1,
-                    value=default_prompt,
-                    label='Prompt'
+                outpaint_prompts = gr.Dataframe(
+                    type="array",
+                    headers=["outpaint steps", "prompt"],
+                    datatype=["number", "str"],
+                    row_count=1,
+                    col_count=(2, "fixed"),
+                    value=[[0, default_prompt]],
+                    wrap=True
                 )
 
                 outpaint_negative_prompt = gr.Textbox(
@@ -185,7 +190,7 @@ def zoom_app():
                     )
                     init_image = gr.Image(type="pil")
                 generate_btn = gr.Button(value='Generate video')
-        
+
             with gr.Column():
                 output_image = gr.Video(label='Output', format="mp4").style(
                     width=512, height=512)
@@ -194,7 +199,7 @@ def zoom_app():
             fn=zoom,
             inputs=[
                 model_id,
-                outpaint_prompt,
+                outpaint_prompts,
                 outpaint_negative_prompt,
                 outpaint_steps,
                 guidance_scale,