add output_dir

animate.py

@@ -1,5 +1,6 @@
 import os
 import sys
+from pathlib import Path
 
 import cv2
 import mediapy
@@ -10,7 +11,6 @@ from image_tools.sizes import resize_and_crop
 from moviepy.editor import CompositeVideoClip
 from moviepy.editor import VideoFileClip as vfc
 from PIL import Image
-from pathlib import Path
 
 
 # get key positions at which frame needs to be generated
@@ -21,11 +21,11 @@ def list_of_positions(num_contours, num_frames=100):
     return positions
 
 
-def contourfinder(image1, image2, text=None, num_frames=100):
+def contourfinder(image1, image2, text=None, num_frames=100, output_dir="temp"):
     # Create two blank pages to write into
     # I just hardcoded 1024*1024 as the size, ideally this should be np.shape(image1)
-    blank=np.zeros(np.shape(image1), dtype="uint8")
-    blank2=np.zeros(np.shape(image2), dtype="uint8") 
+    blank = np.zeros(np.shape(image1), dtype="uint8")
+    blank2 = np.zeros(np.shape(image2), dtype="uint8")
     # Threshold and contours for image 1 and 2
     threshold = cv2.Canny(image=image1, threshold1=100, threshold2=200)
     contours, hierarchies = cv2.findContours(
@@ -39,13 +39,22 @@ def contourfinder(image1, image2, text=None, num_frames=100):
 
     # Initialize three empty videos
     vid1 = cv2.VideoWriter(
-        "vid1.mp4", cv2.VideoWriter_fourcc(*"mp4v"), 24, threshold.shape
+        Path(output_dir / "vid1.mp4"),
+        cv2.VideoWriter_fourcc(*"mp4v"),
+        24,
+        threshold.shape,
     )
     vid2 = cv2.VideoWriter(
-        "vid2.mp4", cv2.VideoWriter_fourcc(*"mp4v"), 24, threshold.shape
+        Path(output_dir / "vid2.mp4"),
+        cv2.VideoWriter_fourcc(*"mp4v"),
+        24,
+        threshold.shape,
     )
     text_vid = cv2.VideoWriter(
-        "text_vid.mp4", cv2.VideoWriter_fourcc(*"mp4v"), 10, threshold.shape
+        Path(output_dir / "text_video.mp4"),
+        cv2.VideoWriter_fourcc(*"mp4v"),
+        10,
+        threshold.shape,
     )
 
     # Get positions
@@ -62,10 +71,9 @@ def contourfinder(image1, image2, text=None, num_frames=100):
             frames.append(blank)
             # Complile to video
             vid1.write(blank)
-            
+
     vid1.release()
-    full_dir_vid_1 = Path("vid1.mp4").resolve().as_posix()
-    clip1 = vfc(full_dir_vid_1)
+    clip1 = vfc(Path(output_dir / "vid1.mp4"))
     positions = list_of_positions((len(contours2)))
 
     for i in range(0, len(contours2)):
@@ -78,8 +86,7 @@ def contourfinder(image1, image2, text=None, num_frames=100):
             vid2.write(blank2)
 
     vid2.release()
-    full_dir_vid_2 = Path("vid2.mp4").resolve().as_posix()
-    clip3 = vfc(full_dir_vid_2)
+    clip3 = vfc(Path(output_dir / "vid2.mp4"))
 
     # Next is the text vid
 
@@ -105,7 +112,7 @@ def contourfinder(image1, image2, text=None, num_frames=100):
         def text_frames(text, image, org):
             spacing = 55  # spacing between letters
             blink = image
-            cv2.imwrite(f"blink.png", blink)
+            cv2.imwrite(Path(output_dir / "blink.png"), blink)
             for i in range(0, len(text) - 1):
 
                 text_vid.write(blink)
@@ -123,7 +130,7 @@ def contourfinder(image1, image2, text=None, num_frames=100):
                 print(org)
 
                 # Displaying the image
-                cv2.imwrite(f"text_im{i}.png", image)
+                cv2.imwrite(Path(output_dir / f"text_im{i}.png"), image)
 
                 # Complile to video
                 text_vid.write(image)
@@ -159,10 +166,7 @@ def resize(width, img):
     return img
 
 
-
-
-
-def resize_img(img1, img2):
+def resize_img(img1, img2, output_dir):
     img_target_size = Image.open(img1)
     img_to_resize = resize_and_crop(
         img2,
@@ -172,10 +176,12 @@ def resize_img(img1, img2):
         ),  # set width and height to match cv2_images[0]
         crop_origin="middle",
     )
-    img_to_resize.save("resized_img2.png")
+    img_to_resize.save(Path(output_dir / "resized_img2.png"))
 
 
-def get_video_frames(images, times_to_interpolate=6, model_name_index=0):
+def get_video_frames(
+    images, vid_output_dir="temp", times_to_interpolate=6, model_name_index=0
+):
     frame1 = images[0]
     frame2 = images[1]
 
@@ -185,11 +191,14 @@ def get_video_frames(images, times_to_interpolate=6, model_name_index=0):
     frame1 = resize(256, frame1)
     frame2 = resize(256, frame2)
 
-    frame1.save("test1.png")
-    frame2.save("test2.png")
+    frame1.save(Path(vid_output_dir / "test1.png"))
+    frame2.save(Path(vid_output_dir / "test2.png"))
 
-    resize_img("test1.png", "test2.png")
-    input_frames = ["test1.png", "resized_img2.png"]
+    resize_img("test1.png", "test2.png", vid_output_dir)
+    input_frames = [
+        Path(vid_output_dir / "test1.png"),
+        Path(vid_output_dir / "resized_img2.png"),
+    ]
 
     frames = list(
         util.interpolate_recursively_from_files(
@@ -200,10 +209,15 @@ def get_video_frames(images, times_to_interpolate=6, model_name_index=0):
 
 
 def create_mp4_with_audio(frames, cv2_images, duration, audio, output_path):
-    temp_vid_path = "TEMP.mp4"
+    vid_output_dir = output_path.parent
+    temp_vid_path = Path(vid_output_dir / "TEMP.mp4")
     mediapy.write_video(temp_vid_path, frames, fps=5)
-    print(f"TYPES....{type(cv2_images[0])},{type(cv2_images[1])} SHAPES{cv2_images[0].shape} Img {cv2_images[0]}")
-    clip1, clip3 = contourfinder(cv2_images[0], cv2_images[1])  # has a third text option
+    print(
+        f"TYPES....{type(cv2_images[0])},{type(cv2_images[1])} SHAPES{cv2_images[0].shape} Img {cv2_images[0]}"
+    )
+    clip1, clip3 = contourfinder(
+        cv2_images[0], cv2_images[1]
+    )  # has a third text option
 
     # Use open CV and moviepy code
     # So we move from open CV video 1 to out.mp4 to open CV video2
@@ -212,7 +226,7 @@ def create_mp4_with_audio(frames, cv2_images, duration, audio, output_path):
     clip3 = clip3.set_start((clip1.duration - 0.5) + (clip2.duration)).crossfadein(2)
 
     new_clip = CompositeVideoClip([clip1, clip2, clip3])
-    new_clip.audio = audio
+    new_clip.audio = audio  # Naviely append audio without considering the length of the video, could be a problem, no idea, but it works, so I'm not touching it
     new_clip.set_duration(duration)
     new_clip.write_videofile(output_path, audio_codec="aac")
     return output_path

app.py

@@ -1,9 +1,10 @@
 import os
-import sys
 import shutil
 import subprocess
+import sys
 import uuid
 from pathlib import Path
+
 import gradio as gr
 from moviepy.editor import AudioFileClip
 
@@ -11,7 +12,6 @@ output_dir = Path("temp/").absolute()
 output_dir.mkdir(exist_ok=True, parents=True)
 
 
-
 class SpotifyApi:
     spotify_directory = Path("spotify")
     final_directory = output_dir
@@ -77,7 +77,10 @@ def process_inputs(
 
 
 def animate_images(image_paths: list[str], audio_input: AudioInput) -> str:
-    from animate import create_mp4_with_audio, get_video_frames # Only import after git clone and when necessary takes loooong
+    from animate import (  # Only import after git clone and when necessary takes loooong
+        create_mp4_with_audio,
+        get_video_frames,
+    )
 
     # Generate a random folder name and change directories to there
     foldername = str(uuid.uuid4())[:8]
@@ -87,7 +90,7 @@ def animate_images(image_paths: list[str], audio_input: AudioInput) -> str:
     audio_clip = audio_clip.subclip(
         audio_input.start_time, audio_input.start_time + audio_input.run_for
     )
-    video_frames, cv2_images = get_video_frames(image_paths)
+    video_frames, cv2_images = get_video_frames(image_paths, vid_output_dir)
     path = Path(vid_output_dir / "output_final.mp4").as_posix()
     return create_mp4_with_audio(
         video_frames, cv2_images, audio_clip.duration, audio_clip, path
@@ -112,12 +115,19 @@ iface = gr.Interface(
     outputs="video",
 )
 
-if __name__ == '__main__':
-    subprocess.call(["git", "clone", "https://github.com/google-research/frame-interpolation", "frame_interpolation"]) # install frame_interplation I guess
+if __name__ == "__main__":
+    subprocess.call(
+        [
+            "git",
+            "clone",
+            "https://github.com/google-research/frame-interpolation",
+            "frame_interpolation",
+        ]
+    )  # install frame_interplation I guess
     sys.path.append("frame_interpolation")
 
     # My installs
     os.chdir(
         output_dir
     )  # change working directory to output_dir because the hf spaces model has no option to specify output directory ¯\_(ツ)_/¯
-    iface.launch(share=True)
+    iface.launch(share=True)