summaryrefslogtreecommitdiffstats
path: root/Scripts
diff options
context:
space:
mode:
Diffstat (limited to 'Scripts')
-rw-r--r--Scripts/emotes.py16
-rw-r--r--Scripts/emotes_v2.py109
-rw-r--r--Scripts/generate_shader.py6
-rw-r--r--Scripts/osc_ctrl.py10
-rw-r--r--Scripts/transcribe.py33
5 files changed, 150 insertions, 24 deletions
diff --git a/Scripts/emotes.py b/Scripts/emotes.py
index 0a4ed01..6ae0930 100644
--- a/Scripts/emotes.py
+++ b/Scripts/emotes.py
@@ -111,16 +111,28 @@ def addImageToTexture(tex: Image, img_path: str, x: int, y:int):
def parseArgs():
parser = argparse.ArgumentParser()
parser.add_argument("--texture_path", type=str, help="Path to save the generated texture.")
+ parser.add_argument("--rows", type=str, help="The number of rows on the board")
+ parser.add_argument("--cols", type=str, help="The number of columns on the board")
args = parser.parse_args()
- if not args.texture_path:
- args.texture_path = "img_texture.png"
+ if not args.texture_path or not args.rows or not args.cols:
+ print("--texture_path, --rows, --cols required", file=sys.stderr)
+ sys.exit(1)
return args
if __name__ == "__main__":
args = parseArgs()
+ rows = int(args.rows)
+ cols = int(args.cols)
+ # board is this much wider than tall
+ board_aspect_ratio = 2
+ # each cell a square divided into `rows`x`cols` is this much wider than tall
+ cell_aspect_ratio = rows / cols
+ # each cell is this much wider than tall
+ board_cell_aspect_ratio = board_aspect_ratio * cell_aspect_ratio
+
tex = openTexture(args.texture_path)
for i in range(0, len(IMG_TEX_DATA)):
filename = IMG_TEX_DATA[i][0]
diff --git a/Scripts/emotes_v2.py b/Scripts/emotes_v2.py
new file mode 100644
index 0000000..195e116
--- /dev/null
+++ b/Scripts/emotes_v2.py
@@ -0,0 +1,109 @@
+#!/usr/bin/env python3
+
+import argparse
+import os
+import pickle
+import sys
+
+from math import floor
+from PIL import Image
+from typing import Any, Dict, List, Tuple
+
+# The character range [0x0000, 0xDFFF] is reserved for text.
+# The range [0xE000, infinity) is left over for emotes.
+EMOTES_LETTER_OFFSET = 0xE000
+
+def superimpose_image(base_img: Image, overlay_img: Image, position: Tuple[int, int]) -> Image:
+ base_img.paste(overlay_img, position, overlay_img)
+ return base_img
+
+def i_to_pos(i, sm_wd, sm_ht, big_wd, big_ht) -> Tuple[int, int]:
+ x = i * sm_wd % big_wd
+ row = floor((i * sm_wd) / big_wd)
+ y = row * sm_ht
+ return int(x), int(y)
+
+def get_images_from_directory(directory_path: str) -> List[Tuple[Any, str]]:
+ images = []
+ for filename in os.listdir(directory_path):
+ file_path = os.path.join(directory_path, filename)
+ if os.path.isfile(file_path) and file_path.endswith(".png"):
+ image = Image.open(file_path)
+ name = os.path.basename(filename).split('.')[0]
+ images.append((image, name))
+ return images
+
+def split_resized_image(img, wd: int, ht: int) -> List[Any]:
+ aspect_ratio = img.width / img.height
+ width = int(ht * aspect_ratio)
+ img = img.resize((width, ht))
+
+ split_images = []
+ for i in range(0, img.width, wd):
+ split_image = img.crop((i, 0, i + wd, ht))
+ split_images.append(split_image)
+
+ return split_images
+
+print(i_to_pos(0, 5, 10, 10, 20))
+print(i_to_pos(1, 5, 10, 10, 20))
+print(i_to_pos(2, 5, 10, 10, 20))
+print(i_to_pos(3, 5, 10, 10, 20))
+
+def resize_image_with_aspect_ratio(img: Image, aspect_ratio: float) -> Image:
+ original_width, original_height = img.size
+ new_width = int(original_height * aspect_ratio)
+ new_height = original_height
+ return img.resize((new_width, new_height))
+
+def resize_image_to_height(img: Image, height: int) -> Image:
+ aspect_ratio = img.width / img.height
+ new_width = int(height * aspect_ratio)
+ return img.resize((new_width, height))
+
+class EmotesState:
+ def __init__(self):
+ self.bits = {}
+
+ def load(self, pickle_path):
+ with open(pickle_path, 'rb') as f:
+ self.bits = pickle.load(f)
+
+if __name__ == "__main__":
+ parser = argparse.ArgumentParser()
+ parser.add_argument("dir", type=str, help="directory to get images from")
+ parser.add_argument("board_aspect_ratio", help="aspect ratio of a cell in the board")
+ parser.add_argument("texture_aspect_ratio", help="aspect ratio of a cell in the texture")
+ parser.add_argument("tex_path", type=str, help="path to save the texture to")
+ parser.add_argument("pickle_path", type=str, help="path to save the texture index to")
+ args = parser.parse_args()
+
+ directory_path = args.dir
+ board_aspect_ratio = int(args.board_aspect_ratio)
+ texture_aspect_ratio = int(args.texture_aspect_ratio)
+
+ base_img = Image.new("RGBA", (4096, 4096), (0, 0, 0, 0))
+ images_and_filenames = get_images_from_directory(directory_path)
+ i = 0
+ bits = {} # Dict[str, List[int]]
+ for img, filename in images_and_filenames:
+ print(f"Adding {filename}")
+ img = resize_image_with_aspect_ratio(img, board_aspect_ratio)
+ img = resize_image_to_height(img, 1024)
+ img_fragments = split_resized_image(img, int(1024 / texture_aspect_ratio), 1024)
+ img_bits = [] # List[int]
+ for img_fragment in img_fragments:
+ i = i + 1
+ img_pos = i_to_pos(i,
+ 1024 / texture_aspect_ratio, 1024,
+ 4096, 4096)
+ print(f"{img_pos}")
+ superimpose_image(base_img, img_fragment, img_pos)
+ img_bits.append(EMOTES_LETTER_OFFSET + i)
+ emote_name = os.path.basename(filename).split('.')[0]
+ print(f"{emote_name} -> {img_bits}")
+ bits[emote_name] = img_bits
+ base_img.save(args.tex_path)
+ with open(args.pickle_path, 'wb') as f:
+ pickle.dump(bits, f)
+
diff --git a/Scripts/generate_shader.py b/Scripts/generate_shader.py
index 9ff0bc3..cf113ec 100644
--- a/Scripts/generate_shader.py
+++ b/Scripts/generate_shader.py
@@ -64,8 +64,8 @@ def generateCgConstants(nbytes: int, nrows: int, ncols: int, prefix: str = "") -
# case 1:
# ...
#
-# res |= ((int) _Letter_Row00_Col00_Byte0) << (0 * 8);
-# res |= ((int) _Letter_Row00_Col00_Byte1) << (1 * 8);
+# res |= ((int) round(_Letter_Row00_Col00_Byte0)) << (0 * 8);
+# res |= ((int) round(_Letter_Row00_Col00_Byte1)) << (1 * 8);
# continue;
# }
# }
@@ -84,7 +84,7 @@ def generateLetterAccessor(nbytes: int, nrows: int, ncols: int, prefix: str = ""
lines.append(prefix + " case {}:".format(col))
for byte in range(0, nbytes):
param_name = generate_utils.getShaderParamByRowColByte(row, col, byte)
- lines.append(prefix + " res |= ((int) {}) << ({} * 8);".format(param_name, byte))
+ lines.append(prefix + " res |= ((int) round({})) << ({} * 8);".format(param_name, byte))
lines.append(prefix + " return res;")
lines.append(prefix + " default:")
lines.append(prefix + " return 0;")
diff --git a/Scripts/osc_ctrl.py b/Scripts/osc_ctrl.py
index 93b236b..750059f 100644
--- a/Scripts/osc_ctrl.py
+++ b/Scripts/osc_ctrl.py
@@ -1,11 +1,13 @@
#!/usr/bin/env python3
-import argparse
+from emotes_v2 import EmotesState
from generate_utils import config
-import generate_utils
+from math import ceil
from paging import MultiLinePager
from pythonosc import udp_client
-from math import ceil
+
+import argparse
+import generate_utils
import time
# Based on a couple experiments, this seems like about as fast as we can go
@@ -87,7 +89,7 @@ def updateRegion(client, region_idx, letter_encoded):
# Sends one slice of `msg` to the board then returns. Slices are sent
# in FIFO order; e.g., the most recently spoken words are sent last.
# Returns True if done paging, False otherwise.
-def pageMessage(osc_state: OscState, msg: str) -> bool:
+def pageMessage(osc_state: OscState, msg: str, estate: EmotesState) -> bool:
msg_slice, slice_idx = osc_state.pager.getNextSlice(msg)
if slice_idx == -1:
return True
diff --git a/Scripts/transcribe.py b/Scripts/transcribe.py
index ee76a0a..3171336 100644
--- a/Scripts/transcribe.py
+++ b/Scripts/transcribe.py
@@ -1,28 +1,23 @@
#!/usr/bin/env python3
+from datetime import datetime
+from emotes_v2 import EmotesState
+from functools import partial
+from playsound import playsound
+
import argparse
import copy
-from datetime import datetime
import os
import osc_ctrl
-from functools import partial
import generate_utils
-# python3 -m pip install pyaudio
-# License: MIT.
import pyaudio
import numpy as np
-# python3 -m pip install playsound==1.2.2
-# License: MIT.
-from playsound import playsound
import steamvr
import string_matcher
import sys
import threading
import time
import wave
-# python3 -m pip install git+https://github.com/openai/whisper.git
-# python3 -m pip install torch -f https://download.pytorch.org/whl/torch_stable.html
-# License: MIT.
import whisper
class Config:
@@ -303,14 +298,14 @@ def transcribeAudio(audio_state, model, use_cpu: bool):
audio_state.transcribe_no_change_count = 0
audio_state.transcribe_sleep_duration = audio_state.transcribe_sleep_duration_min_s
-def sendAudio(audio_state, use_builtin: bool):
+def sendAudio(audio_state, use_builtin: bool, estate: EmotesState):
while audio_state.run_app == True:
text = audio_state.committed_text + " " + audio_state.text
if use_builtin:
ret = osc_ctrl.pageMessageBuiltin(audio_state.osc_state, text)
time.sleep(1.5)
else:
- ret = osc_ctrl.pageMessage(audio_state.osc_state, text)
+ ret = osc_ctrl.pageMessage(audio_state.osc_state, text, estate)
is_paging = (ret == False)
osc_ctrl.indicatePaging(audio_state.osc_state.client, is_paging)
@@ -393,7 +388,7 @@ def readControllerInput(audio_state, enable_local_beep: bool,
# whisper/__init__.py. Examples: tiny, base, small, medium.
def transcribeLoop(mic: str, language: str, model: str,
enable_local_beep: bool, use_cpu: bool, use_builtin: bool,
- button: str):
+ button: str, estate: EmotesState):
audio_state = getMicStream(mic)
audio_state.language = whisper.tokenizer.TO_LANGUAGE_CODE[language]
@@ -410,7 +405,7 @@ def transcribeLoop(mic: str, language: str, model: str,
transcribe_audio_thd.daemon = True
transcribe_audio_thd.start()
- send_audio_thd = threading.Thread(target = sendAudio, args = [audio_state, use_builtin])
+ send_audio_thd = threading.Thread(target = sendAudio, args = [audio_state, use_builtin, estate])
send_audio_thd.daemon = True
send_audio_thd.start()
@@ -459,6 +454,7 @@ if __name__ == "__main__":
parser.add_argument("--cpu", type=int, help="If set to 1, use CPU instead of GPU")
parser.add_argument("--use_builtin", type=int, help="If set to 1, use the text box built into the game.")
parser.add_argument("--button", type=str, help="The controller button used to start/stop transcription. E.g. \"left joystick\"")
+ parser.add_argument("--emotes_pickle", type=str, help="The path to emotes pickle. See emotes_v2.py for details.")
args = parser.parse_args()
if not args.mic:
@@ -482,6 +478,10 @@ if __name__ == "__main__":
print("--button required", file=sys.stderr)
sys.exit(1)
+ if not args.emotes_pickle:
+ print("--emotes_pickle required", file=sys.stderr)
+ sys.exit(1)
+
if args.window_duration_s:
config.MAX_LENGTH_S = int(args.window_duration_s)
@@ -495,11 +495,14 @@ if __name__ == "__main__":
else:
args.use_builtin = False
+ estate = EmotesState()
+ estate.load(args.emotes_pickle)
+
generate_utils.config.BYTES_PER_CHAR = int(args.bytes_per_char)
generate_utils.config.CHARS_PER_SYNC = int(args.chars_per_sync)
generate_utils.config.BOARD_ROWS = int(args.rows)
generate_utils.config.BOARD_COLS = int(args.cols)
transcribeLoop(args.mic, args.language, args.model, args.enable_local_beep,
- args.cpu, args.use_builtin, args.button)
+ args.cpu, args.use_builtin, args.button, estate)
generated by cgit v1.2.3 (git 2.39.1) at 2026-06-01 16:46:38 +0000