misc python

This commit is contained in:
osmarks 2024-02-25 14:26:53 +00:00
parent c6a66dbb8b
commit bb27d89936
4 changed files with 194 additions and 2 deletions

View File

@ -0,0 +1,47 @@
from decimal import Decimal
import itertools
def sum_series(series):
x = Decimal(0)
i = 0
t = 0
l = 0
# ethical
for value in series:
last_x = x
x += value
if x == last_x:
l += 1
else:
t = i
l = 0
if l > (i // 2 + 2): return x
i += 1
def power_series(x, coefs):
def f():
a = Decimal(1)
for coef in coefs:
yield a * coef
a *= x
return sum_series(f())
def derivative(power_series):
next(power_series)
for value, i in zip(power_series, itertools.count(1)):
yield i * value
def cos():
for i in itertools.count(0):
yield Decimal(-1)**i / Decimal(math.factorial(i * 2))
yield 0
def newton(func, x0):
x = x0
while True:
fx, fprimex = power_series(x, func()), power_series(x, derivative(func()))
x = x - fx / fprimex
yield x
for half_pi in newton(cos, 1):
print(half_pi * 2)

93
lesswrong_to_tiktok.py Normal file
View File

@ -0,0 +1,93 @@
import wave, sys
import nltk
from balacoon_tts import TTS
from collections import namedtuple
import struct
from PIL import Image, ImageDraw
import cv2
import numpy
import subprocess
WIDTH = 400
AUDIO = "/tmp/x.wav"
VIDEO = "/tmp/x.avi"
BACKDROP = "/tmp/x.mkv"
OUTPUT = "/tmp/x.mp4"
def render_text(text: str):
render_params = {"font_size": 24}
im = Image.new("RGBA", (WIDTH, WIDTH))
draw = ImageDraw.Draw(im)
loc = [0, 0]
toks = text.split()
toks.reverse()
text_commands = []
while toks:
chunk = []
while draw.textbbox(loc, " ".join(chunk), **render_params)[2] < WIDTH:
if not toks: break
chunk.append(toks.pop())
else: toks.append(chunk.pop())
bbox = draw.textbbox(loc, " ".join(chunk), **render_params)
text_commands.append((tuple(loc), " ".join(chunk)))
loc[1] = bbox[3]
draw.rectangle([0, 0, WIDTH, loc[1]], fill="white")
for loc, text in text_commands:
draw.text(loc, text, fill="black", **render_params)
return im
Pause = namedtuple("Pause", ["length"])
text = open("/home/osmarks/Downloads/seq1.txt").read()
tts = TTS("/home/osmarks/Downloads/en_us_hifi_jets_cpu.addon")
#supported_speakers = tts.get_speakers()
speaker = "6670"
def chunks(text: str) -> list[str | Pause]:
out = []
for line in text.splitlines():
if line:
for sent in nltk.sent_tokenize(line):
out.append(sent)
out.append(Pause(0.5))
out.append(Pause(1))
return out
RATE = tts.get_sampling_rate()
FPS = 30
def wavblank(seconds):
return struct.pack(">h", 0) * round(seconds * RATE / 2)
blank_frame = render_text("")
fourcc = cv2.VideoWriter_fourcc(*"MJPG")
video_writer = cv2.VideoWriter(VIDEO, fourcc, FPS, (WIDTH, WIDTH))
total_dur = 0
with wave.open(AUDIO, "w") as fp:
fp.setparams((1, 2, RATE, 0, "NONE", "NONE"))
for chunk in chunks(text):
if isinstance(chunk, str):
samples = tts.synthesize(chunk, speaker)
image = render_text(chunk)
elif isinstance(chunk, Pause):
samples = wavblank(chunk.length)
image = blank_frame
fp.writeframes(samples)
duration = len(samples) / RATE # what
total_dur += duration
frame = cv2.cvtColor(numpy.array(image), cv2.COLOR_RGBA2BGR)
for _ in range(round(duration * FPS)): video_writer.write(frame)
print(chunk, duration)
video_writer.release()
subprocess.run([
"ffmpeg",
"-i", BACKDROP, "-i", AUDIO, "-i", VIDEO,
"-filter_complex", "overlay=x=200:y=200,format=nv12,hwupload",
"-to", str(total_dur),
"-y",
"-vaapi_device", "/dev/dri/renderD128", "-c:v", "h264_vaapi",
OUTPUT
]).check_returncode()

View File

@ -1,6 +1,6 @@
import os, sys, subprocess, datetime
dt_threshold = datetime.datetime(2022, 11, 27).timestamp()
dt_threshold = datetime.datetime(2023, 6, 16).timestamp()
_, indir, outdir = sys.argv
for x in os.listdir(indir):
@ -19,4 +19,4 @@ for x in os.listdir(indir):
os.rename(inpath, newpath)
break
else:
print("keeping")
print("keeping")

52
rotating_audio.py Normal file
View File

@ -0,0 +1,52 @@
import pathlib
import numpy as np
import wave
import math
import tempfile
import subprocess
import contextlib
cyc = 2
nsamples = 48000 * 60
path = pathlib.Path("~/Downloads/inputs").expanduser()
inputs = []
tempfiles = []
with contextlib.ExitStack() as stack:
for file in path.glob("*"):
mgr = tempfile.NamedTemporaryFile("w")
stack.enter_context(mgr)
tempfiles.append(mgr)
subprocess.run(["ffmpeg", "-i", file, "-ar", "48000", "-ac", "1", "-filter:a", "dynaudnorm=p=1.0:s=5:g=15", "-y", "-f", "wav", mgr.name])
for t in tempfiles:
with wave.open(str(t.name), "r") as w:
assert w.getsampwidth() == 2 and w.getnchannels() == 1 and w.getframerate() == 48000
inputs.append(np.frombuffer(w.readframes(nsamples), np.dtype(np.int16).newbyteorder("<")).astype(float) / 32768)
inputs = np.array(inputs) * (1/len(inputs)) # wrong but close enough
offsets = np.broadcast_to(np.linspace(0, math.pi * 2, num=len(inputs) + 1)[:len(inputs)], (nsamples, len(inputs))).transpose()
phase = np.broadcast_to(np.linspace(0, math.pi * 2 * cyc, num=nsamples), (len(inputs), nsamples)) + offsets
empty = np.zeros_like(phase)
x = np.cos(phase)
y = np.sin(phase)
def writeaudio(file, array):
wr = wave.open(file, "w")
wr.setnchannels(1)
wr.setsampwidth(2)
wr.setframerate(48000)
wr.writeframes((array * 32768).astype("<i2").tobytes())
wr.close()
with tempfile.NamedTemporaryFile("wb") as left:
with tempfile.NamedTemporaryFile("wb") as front:
with tempfile.NamedTemporaryFile("wb") as right:
with tempfile.NamedTemporaryFile("wb") as back:
writeaudio(left, sum(np.where(-x > 0, -x, empty) * inputs))
writeaudio(right, sum(np.where(x > 0, x, empty) * inputs))
writeaudio(front, sum(np.where(y > 0, y, empty) * inputs))
writeaudio(back, sum(np.where(-y > 0, -y, empty) * inputs))
print(front.name)
# layout is technically front left + front right + back left + back right - ignore
subprocess.run(["ffmpeg", "-i", front.name, "-i", right.name, "-i", back.name, "-i", left.name, "-filter_complex", "[0:a][1:a]join=inputs=4:channel_layout=quad[a]", "-map", "[a]", "/tmp/out.opus"]).check_returncode()