auracaster/auracast-translator
3
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

refractoring to use ghe global config.py

Browse Source
This commit is contained in:
pstruebi
2024-12-19 16:28:52 +01:00
parent 5d2f0f4fdf
commit 56b942ce39
6 changed files with 61 additions and 65 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
multilang_translator/backend_controller/broadcaster_config.py
View File

@@ -1,20 +1,16 @@
import serial
import time
import logging as log
import os
import serial
from ..config import SAMPLING_RATE_HZ, LANG_CONFIG
PRESET = f'{SAMPLING_RATE_HZ//1000}_2_1'
BROADCAST_CONFIG = {}
SAMPLING_RATE_KHZ = 16
FRAME_DUR_MS = 10
PRESET = f'{SAMPLING_RATE_KHZ}_2_1'
for i, d in enumerate(list(LANG_CONFIG.items())):
key, val = d
BROADCAST_CONFIG[i] = os.path.basename(val["file"])
if SAMPLING_RATE_KHZ == 8:
BITRATE_KBPS = 16
elif SAMPLING_RATE_KHZ == 16:
BITRATE_KBPS = 32
elif SAMPLING_RATE_KHZ == 24:
BITRATE_KBPS = 48
else:
raise NotImplemented()
def write_to_serial_read_respone(port, cmd, timeout = 2):
# Initialize serial connection
@@ -68,19 +64,12 @@ def gen_broadcast_config_cmd(preset, broadcast_config: dict):
cmds.append(f"nac preset {preset} {ch}")
cmds.append(f"nac broadcast_name broadcast{ch} {ch}")
cmds.append(f"nac file select_play_once {file_name} {ch} 0 0")
cmds.append(f"nac file select_play_once {file_name}.lc3 {ch} 0 0")
cmds.append(f"nac num_bises 1 {ch} 0")
return cmds
# TODO: Advertising interval wird ungelmäßig bei mehr als 3 broadcasts 10ms -> 1s< bei 24kHz sampling rate
BROADCAST_CONFIG = {
0: f"announcement_{SAMPLING_RATE_KHZ}_{FRAME_DUR_MS}_{BITRATE_KBPS}_de.lc3",
1: f"announcement_{SAMPLING_RATE_KHZ}_{FRAME_DUR_MS}_{BITRATE_KBPS}_en.lc3",
2: f"announcement_{SAMPLING_RATE_KHZ}_{FRAME_DUR_MS}_{BITRATE_KBPS}_fr.lc3",
#3: f"announcement_{SAMPLING_RATE_KHZ}_{FRAME_DUR_MS}_{BITRATE_KBPS}_es.lc3",
#4: f"announcement_{SAMPLING_RATE_KHZ}_{FRAME_DUR_MS}_{BITRATE_KBPS}_it.lc3"
}
def broadcaster_config():

30
multilang_translator/config.py
View File

@@ -0,0 +1,30 @@
import os
ANNOUNCEMENT_DIR = os.path.join(os.path.dirname(__file__), 'announcements')
SAMPLING_RATE_HZ = int(16e3)
FRAME_DUR_MS = 10
BITRATE_BPS = int(32e3) # TODO: test 16khz 16kbps
LANG_CONFIG = {
"de": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE_HZ//1000}_{FRAME_DUR_MS}_{BITRATE_BPS//1000}_de",
"tts": 'de_DE-kerstin-low',
},
"en": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE_HZ//1000}_{FRAME_DUR_MS}_{BITRATE_BPS//1000}_en",
"tts": 'en_US-lessac-medium'
},
"fr": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE_HZ//1000}_{FRAME_DUR_MS}_{BITRATE_BPS//1000}_fr",
"tts": 'fr_FR-siwis-medium'
},
# "es": {
# "file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE_HZ//1000}_{FRAME_DUR_MS}_{BITRATE_BPS//1000}_es",
# "tts": 'es_ES-sharvard-medium'
# },
# "it": {
# "file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE_HZ//1000}_{FRAME_DUR_MS}_{BITRATE_BPS//1000}_it",
# "tts": 'it_IT-paola-medium'
# }
}
os.makedirs(ANNOUNCEMENT_DIR, exist_ok=True)

45
multilang_translator/main.py
View File

@@ -19,54 +19,27 @@ from .text_to_speech import text_to_speech, resample
from .backend_controller.broadcaster_config import broadcaster_config
from .backend_controller.broadcaster_play_once import broadcaster_play_file
from .backend_controller.broadcaster_copy_files import copy_to_broadcaster
from .encode import encode_lc3
ANNOUNCEMENT_DIR = os.path.join(os.path.dirname(__file__), 'announcements')
SAMPLING_RATE = int(16e3)
FRAME_DUR_MS = 10
BPS = int(32e3) # TODO test 16khz 16kbps
CONFIG = {
"de": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE//1000}_{FRAME_DUR_MS}_{BPS//1000}_de",
"tts": 'de_DE-kerstin-low',
},
"en": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE//1000}_{FRAME_DUR_MS}_{BPS//1000}_en",
"tts": 'en_US-lessac-medium'
},
"fr": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE//1000}_{FRAME_DUR_MS}_{BPS//1000}_fr",
"tts": 'fr_FR-siwis-medium'
},
"es": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE//1000}_{FRAME_DUR_MS}_{BPS//1000}_es",
"tts": 'es_ES-sharvard-medium'
},
"it": {
"file": f"{ANNOUNCEMENT_DIR}/announcement_{SAMPLING_RATE//1000}_{FRAME_DUR_MS}_{BPS//1000}_it",
"tts": 'it_IT-paola-medium'
}
}
from .config import LANG_CONFIG, BITRATE_BPS, SAMPLING_RATE_HZ, FRAME_DUR_MS
os.makedirs(ANNOUNCEMENT_DIR, exist_ok=True)
def synthesize_resample_encode(text, tts_model, output_file):
audio_dur = text_to_speech.synthesize(text, tts_model, output_file)
resample.resample(output_file, output_file, target_rate=SAMPLING_RATE)
encode_lc3.encode_lc3(output_file, bps=BPS, frame_dur_ms=FRAME_DUR_MS)
resample.resample(output_file, output_file, target_rate=SAMPLING_RATE_HZ)
encode_lc3.encode_lc3(output_file, bps=BITRATE_BPS, frame_dur_ms=FRAME_DUR_MS)
return audio_dur
def translate_from_german_and_encode(text_de):
config = copy(CONFIG)
config = copy(LANG_CONFIG)
base_lang = "de"
file = config[base_lang]["file"]
audio_dur_s = {}
audio_dur_s [base_lang] = synthesize_resample_encode(text_de, config['de']["tts"], f'{file}.wav')
del config[base_lang]
audio_dur_s = {}
for key, val in config.items():
text = llm_translator.translate_de_to_x(key, text_de)
file = val['file']
@@ -78,15 +51,17 @@ def announcement_from_german_text(text_de):
audio_durs = translate_from_german_and_encode(text_de)
# Transfer the files to broadcaster memory
start = time.time()
for val in CONFIG.values():
for val in LANG_CONFIG.values():
copy_to_broadcaster(f'{val["file"]}.lc3')
log.info("Transfering files to broadcaster took %s s", round(time.time() - start, 3))
time.sleep(2)
# Instruct the broadcaster to stream the files
for i, d in enumerate(list(CONFIG.items())):
for i, d in enumerate(list(LANG_CONFIG.items())):
key, val = d
broadcaster_play_file(i, f'{os.path.basename(val["file"])}.lc3')
time.sleep(audio_durs[key] + 1)
time.sleep(audio_durs[key])
log.info("Starting all broadcasts %s s", round(time.time() - start, 3))

4
multilang_translator/translator/test_content.py
View File

@@ -1,2 +1,4 @@
TESTSENTENCE_DE_HELLO = 'Hallo Welt.'
TESTSENTENCE_DE_WAVE_PARTICLE = 'Der Wellen-Teilchen-Dualismus beschreibt die Eigenschaft von Teilchen, sowohl als Wellen auf der Mikroebene zu verhalten und gleichzeitig bestimmte Eigenschaften wie Impuls und Energietrang zu besitzen.'
TESTSENTENCE_DE_BROKER = 'Ein Broker (oder Makler) ist eine Person oder ein Unternehmen, das sich zwischen dem Kauf- und Verkaufsberechtigten einer Wirtschaftsgüter (z.B. Aktien, Optionen, Derivate, Währungen, Rohstoffe usw.) stellt und als Vermittler fungiert. Sein Hauptziel ist es, Transaktionen zu erleichtern und Geld für sich selbst zu verdienen.'
TESTSENTENCE_DE_RAINBOW = 'Der Regenbogen ist ein atmosphärisch-optisches Phänomen, das als kreisbogenförmiges farbiges Lichtband in einer von der Sonne beschienenen Regenwand oder wolke wahrgenommen wird. '
TESTSENTENCE_DE_RAINBOW = 'Der Regenbogen ist ein atmosphärisch-optisches Phänomen, das als kreisbogenförmiges farbiges Lichtband in einer von der Sonne beschienenen Regenwand oder wolke wahrgenommen wird.'