jukebox gradio

app.py

@@ -0,0 +1,103 @@
+from interactive_pipe import Control
+from interactive_pipe.headless.pipeline import HeadlessPipeline
+from interactive_pipe.graphical.gradio_gui import InteractivePipeGradio
+from interactive_pipe import interactive
+from interactive_pipe.data_objects.image import Image
+from interactive_pipe.thirdparty.images_openai_api import ImageFromPrompt
+from pathlib import Path
+import cv2
+
+root = Path(__file__).parent
+img_folder = root/"images"
+audio_folder = root/"audio"
+TRACK = "track"
+IMAGE = ICON = "image"
+CAPTION = "caption"
+PROMPT = "prompt"
+PROMPT_STYLE = "a cute crayon drawing of "
+PROMPT_EXTRA = " on a white background, using a style from a cartoon book for baby kids"
+
+# Audio generated by https://huggingface.co/spaces/facebook/MusicGen
+# Elephant : a short savanah jungle music like the lion's king akuna matata with elephants barking
+# Snail: a short instrumental country song with banjo without lyrics
+# Rabbit: we will rock you by queens but played with ukulele  without lyrics
+TRACK_DICT = {
+    "elephant": {
+        TRACK: audio_folder/"elephant.mp4",
+        CAPTION: "ELEPHANT",
+        PROMPT: "a smiling elephant walking in the sunny yellow savana",
+    },
+    "snail": {
+        TRACK: audio_folder/"snail.mp4",
+        CAPTION: "SNAIL",
+        PROMPT: "a cute yellow and orange snail with two eyes slowly walking on the green grass"
+    },
+    "rabbit": {
+        TRACK: audio_folder/"rabbit.mp4",
+        CAPTION: "RABBIT",
+        PROMPT: "a smiling funny little rabbit in the green grass",
+    },
+    "pause": {
+        TRACK: None,
+        PROMPT: "a cute sleeping cat",
+        CAPTION: "...zzzz"
+    }
+}
+
+for item_name, element in TRACK_DICT.items():
+    TRACK_DICT[item_name][IMAGE] = ImageFromPrompt.generate_image(
+        PROMPT_STYLE + element[PROMPT] + PROMPT_EXTRA,
+        img_folder/(item_name+".png"),
+        size=(512, 512)
+    )
+ICONS = [it[ICON] for key, it in TRACK_DICT.items()]
+
+
+@interactive(
+    song=Control("elephant", list(TRACK_DICT.keys()), icons=ICONS)
+)
+def song_choice(global_params={}, song="elephant"):
+    global_params[TRACK] = song
+
+
+def play_song(global_params={}):
+    song = global_params.get(TRACK, None)
+    first_exec = global_params.get("first_exec", True)
+    if not first_exec:
+        audio_track = TRACK_DICT[song][TRACK]
+        if audio_track is None:
+            global_params["__stop"]()
+        else:
+            global_params["__set_audio"](audio_track)
+            global_params["__play"]()
+    else:
+        global_params["first_exec"] = False
+
+
+def image_choice(global_params={}):
+    song = global_params.get(TRACK, list(TRACK_DICT.keys())[0])
+    img = Image.from_file(TRACK_DICT[song][IMAGE]).data
+    max_height = 300  # Raspberry pi with a 7" touchscreen
+    h, w, _c = img.shape
+    if h > max_height:
+        img = cv2.resize(img, (w*max_height//h, max_height))
+        h, w, _c = img.shape
+    caption = TRACK_DICT[song][CAPTION]
+    global_params["__output_styles"]["img_out"] = {
+        "title": caption
+    }  # discard auto titling
+    return img
+
+
+def sample_pipeline():
+    song_choice()
+    play_song()
+    img_out = image_choice()
+    return [img_out,]
+
+
+if __name__ == '__main__':
+    pip = HeadlessPipeline.from_function(sample_pipeline, cache=False)
+    app = InteractivePipeGradio(
+        pipeline=pip, name="music", size=None, audio=True)
+    app()