How to capture pygame screen?
Question:
How can I capture and save a sequence of images or a video of a pygame screen?
Basically I want to share my game video on youtube. Also, want to make a tutorial.
The game is rendered mainly in a loop:
def main():
while True:
GetInput()
Move()
Shift()
Draw()
With the Draw() function doing all the blit() and stuff before doing the pygame.display.flip()
Answers:
Use pygame.image.save on your screen surface:
window = pygame.display.set_mode(...)
...
pygame.image.save(window, "screenshot.jpeg")
Note that this will slow down your program tremendously. If it is time-based, you may wish to fake the framerate when doing a capture.
The accepted answer said you could save the current screen with the pygame.image.save method
This is a good idea, however you might not want to save the images on the hard drive while the game is running, also pointed out by the question. Instead, you should save the screens in the program and then process them after the game has stopped running.
Here is my code with comments, that only shows the general idea of how a screen recording might work. It uses opencv (with numpy) and pygame but you have to install ffmpeg to convert the images to a video (try ffmpg in the terminal to test). Don’t let the program run too long, because the saving still takes quite a while and is about proportional to the recorded frames. For more efficiency, you could only record every second frame or so.
from contextlib import contextmanager
import os
import time
import cv2
import numpy as np
import pygame as pg
from pygame import surfarray
def pg_to_cv2(cvarray:np.ndarray)->np.ndarray:
cvarray = cvarray.swapaxes(0,1) #rotate
cvarray = cv2.cvtColor(cvarray, cv2.COLOR_RGB2BGR) #RGB to BGR
return cvarray
def timer_wrapper(func):
def inner(*args, **kwargs):
start = time.time()
func(*args, **kwargs)
end = time.time()
#print("Finished:" ,func.__name__ ,end-start)
return end - start
return inner
@contextmanager
def video_writer(*args,**kwargs):
video = cv2.VideoWriter(*args,**kwargs)
try:
yield video
finally:
video.release()
@timer_wrapper
def save_frames(frames: list, average_dt: float|list, file_type: str = "mp4", name: str = "screen_recording"):
if type(average_dt) is list: average_dt = sum(average_dt)/len(average_dt) # force average_dt to be a float
size = frames[0].get_size()
codec_dict={
"avi":'DIVX',
"mp4":'MP4V'
}
codec = cv2.VideoWriter_fourcc(*codec_dict[file_type])
with video_writer(name+"."+file_type, codec, 1000/average_dt, size) as video: # file_name, codec, average_fps, dimensions
for frame in frames:
try:
pg_frame = surfarray.pixels3d(frame) # convert the surface to a np array. Only works with depth 24 or 32, not less
except:
pg_frame = surfarray.array3d(frame) # convert the surface to a np array. Works with any depth
cv_frame = pg_to_cv2(pg_frame) # then convert the np array so it is compatible with opencv
video.write(cv_frame) #write the frame to the video using opencv
def draw_fps(s:pg.Surface,clock:time.Clock):
fps = clock.get_fps()
sysfont.render_to(s,(100,100),str(fps),fgcolor=nice_green)
# initializing globals (colors, fonts, window, etc.)
pg.init()
sysfont = pg.freetype.SysFont(None,40)
BLACK = (0,)*3
nice_green = pg.Color("chartreuse2")
size=(1000, 600)
pg.display.set_caption("Screen Recording")
window = pg.display.set_mode(size)
# this is to save the frames
frames = []
dts = []
clock = pg.time.Clock()
running=True
try:
while running:
dt = clock.tick(60) # aim for ... fps
for event in pg.event.get():
if event.type == pg.QUIT:
running=False
if event.type == pg.KEYDOWN:
if event.key == pg.K_ESCAPE:
running=False
window.fill(BLACK)
draw_fps(window,clock)
window_copy = window.copy() # if we don't copy the window then we will have the same image in all frames at the end
frames.append(window_copy) # We save the current frame together with the time passed between the frames
dts.append(dt)
pg.display.flip()
#if len(frames) >= 100: running = False # uncomment this to stop the game after ... frames for similar results in every run"
finally:
pg.quit()
# At this stage, the game ended and the cleanup process can start. For this we convert the frames to opencv images
# Only then we will write the video to the hard drive (That is what makes the save so slow).
# General information about the recording
frame_num = len(frames)
dt_sum = sum(dts)
average_dt = dt_sum/frame_num
# This is only an approximation:
# for each frame we have width * height many pixels -> frame_num * width * height
# A Surface needs get_bytesize() many bytes per pixel (In this case 4 bytes, because we set the depth of the display to 32 bits)
memory_usage_approx = frame_num * size[0] * size[1] * frames[0].get_bytesize() #https://www.pygame.org/docs/ref/surface.html#pygame.Surface.get_bytesize
print("Total time:" , dt_sum/1000,"s")
print("Average time per frame:" , average_dt,"ms")
print("Number of frames:", frame_num)
print("Memory usage approximation" , memory_usage_approx/1000, "KB")
args = (frames,dts,"avi","screen_recording")
time_for_save = save_frames(*args)
file_name = args[3]+"."+args[2]
video_memory_usage = os.path.getsize(file_name)
print("Video memory usage:" , video_memory_usage/1000, "KB")
with open("test.txt", "a") as f:
print("Total time:" , dt_sum/1000,"snNumber of frames:", frame_num,"nSize:",size,"nTime for save:",time_for_save,"snSaved in file:",file_name,file=f)
print("_"*100,file=f)
Or you just use a lib like Pygame Recorder.
x3 = pygame.surfarray.pixels3d(screen)
x3 = x3[:,:,::-1]
I found a cool way;
you can use
x3 = pygame.surfarray.pixels3d(window)
to get all the pixels on any surface or the screen (the window variable)!
You can use this in things like NumPy where you can use the code
array = numpy.uint8(x3)
to get the image of the surface as a NumPy array and then
im = PIL.Image.fromarray(array)
to make it a Pillow image, if you want. Then you can show it with a simple im.show() or just do whatever with it.
How can I capture and save a sequence of images or a video of a pygame screen?
Basically I want to share my game video on youtube. Also, want to make a tutorial.
The game is rendered mainly in a loop:
def main():
while True:
GetInput()
Move()
Shift()
Draw()
With the Draw() function doing all the blit() and stuff before doing the pygame.display.flip()
Use pygame.image.save on your screen surface:
window = pygame.display.set_mode(...)
...
pygame.image.save(window, "screenshot.jpeg")
Note that this will slow down your program tremendously. If it is time-based, you may wish to fake the framerate when doing a capture.
The accepted answer said you could save the current screen with the pygame.image.save method
This is a good idea, however you might not want to save the images on the hard drive while the game is running, also pointed out by the question. Instead, you should save the screens in the program and then process them after the game has stopped running.
Here is my code with comments, that only shows the general idea of how a screen recording might work. It uses opencv (with numpy) and pygame but you have to install ffmpeg to convert the images to a video (try ffmpg in the terminal to test). Don’t let the program run too long, because the saving still takes quite a while and is about proportional to the recorded frames. For more efficiency, you could only record every second frame or so.
from contextlib import contextmanager
import os
import time
import cv2
import numpy as np
import pygame as pg
from pygame import surfarray
def pg_to_cv2(cvarray:np.ndarray)->np.ndarray:
cvarray = cvarray.swapaxes(0,1) #rotate
cvarray = cv2.cvtColor(cvarray, cv2.COLOR_RGB2BGR) #RGB to BGR
return cvarray
def timer_wrapper(func):
def inner(*args, **kwargs):
start = time.time()
func(*args, **kwargs)
end = time.time()
#print("Finished:" ,func.__name__ ,end-start)
return end - start
return inner
@contextmanager
def video_writer(*args,**kwargs):
video = cv2.VideoWriter(*args,**kwargs)
try:
yield video
finally:
video.release()
@timer_wrapper
def save_frames(frames: list, average_dt: float|list, file_type: str = "mp4", name: str = "screen_recording"):
if type(average_dt) is list: average_dt = sum(average_dt)/len(average_dt) # force average_dt to be a float
size = frames[0].get_size()
codec_dict={
"avi":'DIVX',
"mp4":'MP4V'
}
codec = cv2.VideoWriter_fourcc(*codec_dict[file_type])
with video_writer(name+"."+file_type, codec, 1000/average_dt, size) as video: # file_name, codec, average_fps, dimensions
for frame in frames:
try:
pg_frame = surfarray.pixels3d(frame) # convert the surface to a np array. Only works with depth 24 or 32, not less
except:
pg_frame = surfarray.array3d(frame) # convert the surface to a np array. Works with any depth
cv_frame = pg_to_cv2(pg_frame) # then convert the np array so it is compatible with opencv
video.write(cv_frame) #write the frame to the video using opencv
def draw_fps(s:pg.Surface,clock:time.Clock):
fps = clock.get_fps()
sysfont.render_to(s,(100,100),str(fps),fgcolor=nice_green)
# initializing globals (colors, fonts, window, etc.)
pg.init()
sysfont = pg.freetype.SysFont(None,40)
BLACK = (0,)*3
nice_green = pg.Color("chartreuse2")
size=(1000, 600)
pg.display.set_caption("Screen Recording")
window = pg.display.set_mode(size)
# this is to save the frames
frames = []
dts = []
clock = pg.time.Clock()
running=True
try:
while running:
dt = clock.tick(60) # aim for ... fps
for event in pg.event.get():
if event.type == pg.QUIT:
running=False
if event.type == pg.KEYDOWN:
if event.key == pg.K_ESCAPE:
running=False
window.fill(BLACK)
draw_fps(window,clock)
window_copy = window.copy() # if we don't copy the window then we will have the same image in all frames at the end
frames.append(window_copy) # We save the current frame together with the time passed between the frames
dts.append(dt)
pg.display.flip()
#if len(frames) >= 100: running = False # uncomment this to stop the game after ... frames for similar results in every run"
finally:
pg.quit()
# At this stage, the game ended and the cleanup process can start. For this we convert the frames to opencv images
# Only then we will write the video to the hard drive (That is what makes the save so slow).
# General information about the recording
frame_num = len(frames)
dt_sum = sum(dts)
average_dt = dt_sum/frame_num
# This is only an approximation:
# for each frame we have width * height many pixels -> frame_num * width * height
# A Surface needs get_bytesize() many bytes per pixel (In this case 4 bytes, because we set the depth of the display to 32 bits)
memory_usage_approx = frame_num * size[0] * size[1] * frames[0].get_bytesize() #https://www.pygame.org/docs/ref/surface.html#pygame.Surface.get_bytesize
print("Total time:" , dt_sum/1000,"s")
print("Average time per frame:" , average_dt,"ms")
print("Number of frames:", frame_num)
print("Memory usage approximation" , memory_usage_approx/1000, "KB")
args = (frames,dts,"avi","screen_recording")
time_for_save = save_frames(*args)
file_name = args[3]+"."+args[2]
video_memory_usage = os.path.getsize(file_name)
print("Video memory usage:" , video_memory_usage/1000, "KB")
with open("test.txt", "a") as f:
print("Total time:" , dt_sum/1000,"snNumber of frames:", frame_num,"nSize:",size,"nTime for save:",time_for_save,"snSaved in file:",file_name,file=f)
print("_"*100,file=f)
Or you just use a lib like Pygame Recorder.
x3 = pygame.surfarray.pixels3d(screen)
x3 = x3[:,:,::-1]
I found a cool way;
you can use
x3 = pygame.surfarray.pixels3d(window)
to get all the pixels on any surface or the screen (the window variable)!
You can use this in things like NumPy where you can use the code
array = numpy.uint8(x3)
to get the image of the surface as a NumPy array and then
im = PIL.Image.fromarray(array)
to make it a Pillow image, if you want. Then you can show it with a simple im.show() or just do whatever with it.