Created
January 7, 2020 00:41
-
-
Save lyleaf/a094a514c6cb15b36aead0b8c9cfc394 to your computer and use it in GitHub Desktop.
Create animation which is synced up to the music tempo.
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import cv2 | |
import librosa | |
import librosa.display | |
import wave | |
import pyaudio | |
from cv2 import VideoWriter, VideoWriter_fourcc | |
VIDEO_PATH = 'Demo_Full_1.mp4' | |
video = cv2.VideoCapture(VIDEO_PATH) | |
# Read a video to get FPS | |
fps = video.get(cv2.CAP_PROP_FPS) | |
print('Image frames per second is %d' % fps) | |
video.release() | |
# Create audio mp3 from mp4 | |
# ffmpeg -i holdon.mp4 holdon.mp3 not the same length | |
# ffmpeg -i holdon.mp4 -async 1 holdon.wav | |
AUDIO_PATH = 'Demo_Full_1.wav' # difference between wav and mp3 | |
seconds = 5 | |
y, sr = librosa.load(AUDIO_PATH) | |
print(int(len(y))) | |
print('Sample rate %d' % sr) | |
time = int(len(y)/sr) | |
print('Song time %d' % time) | |
interval = [] | |
for i in range(time): | |
sample_frames = sr * seconds | |
sample = y[i*sr:i*sr+sample_frames] | |
tempo, beat_frames = librosa.beat.beat_track(y=sample, sr=sr) #y should be audio time series np.ndarray [shape=(n,)] or None | |
#print('Estimated tempo: {:.2f} beats per minute'.format(tempo)) | |
interval.append(60/tempo) | |
print(interval) | |
downtempo = [] #in seconds | |
downtempo.append(5) | |
while(downtempo[-1] <= time): | |
downtempo.append(interval[int(downtempo[-1])-5]+downtempo[-1]) | |
print(downtempo) | |
uptempo = [] | |
for i in range(len(downtempo)-1): | |
uptempo.append((downtempo[i]+downtempo[i+1])/2) | |
tempo = [] | |
for i in range(len(uptempo)): | |
tempo.append(downtempo[i]) | |
tempo.append(uptempo[i]) | |
tempo_frames = np.multiply(tempo,fps).astype(int) | |
print(tempo_frames) | |
# Create output video with animation | |
width = 1280 | |
height = 720 | |
FPS = fps | |
radius = 150 | |
paint_h = int(height/2) | |
paint_x = int(width/2) | |
fourcc = VideoWriter_fourcc(*'MP42') | |
video = VideoWriter('./Demo_Full_1_noises.avi', fourcc, float(FPS), (width, height)) | |
index = 0 | |
black = False | |
for timestamp in range(int(FPS)*time): | |
frame = np.random.randint(0, 256, | |
(height, width, 3), | |
dtype=np.uint8) | |
if ((index+2 < len(tempo_frames)-1) and (timestamp >= tempo_frames[index]) and (timestamp < tempo_frames[index+2])): | |
print(timestamp) | |
if (timestamp < tempo_frames[index+1]): | |
cv2.circle(frame, (paint_x, paint_h), radius, (0, 0, 0), -1) | |
print('add frame') | |
elif ((index+2 < len(tempo_frames)-1) and (timestamp >= tempo_frames[index+2])): | |
index = index+2 | |
video.write(frame) | |
video.release() | |
# ffmpeg -i Demo_Full_1_noises.avi NOISES.mp4 |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment