rixty/KoboldAI-Client
1
1
Fork 0
mirror of https://github.com/KoboldAI/KoboldAI-Client.git synced 2025-06-05 21:59:24 +02:00
Code Issues Projects Releases Wiki Activity

Merge branch 'UI2' of https://github.com/ebolam/KoboldAI into ui2-story-review

Browse Source
This commit is contained in:
somebody
2022-12-10 10:56:11 -06:00
parent 72c9ff6763 f21c3c7c09
commit 3f037cda1f
2 changed files with 18 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
aiserver.py
View File

@@ -1269,14 +1269,6 @@ def loadsettings():
with open("settings/" + getmodelname().replace('/', '_') + ".v2_settings", "r") as file: with open("settings/" + getmodelname().replace('/', '_') + ".v2_settings", "r") as file:
getattr(koboldai_vars, "_model_settings").from_json(file.read()) getattr(koboldai_vars, "_model_settings").from_json(file.read())
file.close()
if(path.exists(get_config_filename())):
# Read file contents into JSON object
file = open(get_config_filename(), "r")
js = json.load(file)
processsettings(js)
file.close()
def processsettings(js): def processsettings(js):
# Copy file contents to vars # Copy file contents to vars

28
koboldai_settings.py
View File

@@ -18,6 +18,9 @@ serverstarted = False
queue = None queue = None
multi_story = False multi_story = False
if importlib.util.find_spec("tortoise") is not None:
from tortoise import api
from tortoise.utils.audio import load_voices
def clean_var_for_emit(value): def clean_var_for_emit(value):
if isinstance(value, KoboldStoryRegister) or isinstance(value, KoboldWorldInfo): if isinstance(value, KoboldStoryRegister) or isinstance(value, KoboldWorldInfo):
@@ -932,7 +935,8 @@ class story_settings(settings):
if self.story_id == j["story_id"]: if self.story_id == j["story_id"]:
break break
except FileNotFoundError: except FileNotFoundError:
raise FileNotFoundError("Malformed save file: Missing story.json") logger.error(f"Malformed save file: Missing story.json in {self.save_paths.base}. Populating it with new data.")
break
disambiguator += 1 disambiguator += 1
self.save_paths.base = os.path.join("stories", save_name + (f" ({disambiguator})" if disambiguator else "")) self.save_paths.base = os.path.join("stories", save_name + (f" ({disambiguator})" if disambiguator else ""))
@@ -950,8 +954,11 @@ class story_settings(settings):
logger.info("Migrating v2 save") logger.info("Migrating v2 save")
with open(v2_path, "r") as file: with open(v2_path, "r") as file:
v2j = json.load(file) v2j = json.load(file)
assert v2j["story_id"] == self.story_id
shutil.move(v2_path, os.path.join(self.save_paths.base, ".v2_old.json")) if v2j["story_id"] == self.story_id:
shutil.move(v2_path, os.path.join(self.save_paths.base, ".v2_old.json"))
else:
logger.warning(f"Story mismatch in v2 migration. Existing file had story id {v2j['story_id']} but we have {self.story_id}")
with open(self.save_paths.story, "w") as file: with open(self.save_paths.story, "w") as file:
file.write(self.to_json()) file.write(self.to_json())
@@ -2001,16 +2008,16 @@ class KoboldStoryRegister(object):
def create_wave_slow(self, make_audio_queue_slow): def create_wave_slow(self, make_audio_queue_slow):
import pydub import pydub
sample_rate = 24000 sample_rate = 24000
speaker = 'train_daws'
if self.tortoise is None: if self.tortoise is None:
try: self.tortoise=api.TextToSpeech()
from tortoise import api
self.tortoise=api.TextToSpeech()
except:
self.tortoise = False
if self.tortoise is not False: if importlib.util.find_spec("tortoise") is not None:
voice_samples, conditioning_latents = load_voices([speaker])
while not make_audio_queue_slow.empty(): while not make_audio_queue_slow.empty():
start_time = time.time()
(text, filename) = make_audio_queue_slow.get() (text, filename) = make_audio_queue_slow.get()
text_length = len(text)
logger.info("Creating audio for {}".format(os.path.basename(filename))) logger.info("Creating audio for {}".format(os.path.basename(filename)))
if text.strip() == "": if text.strip() == "":
shutil.copy("data/empty_audio.ogg", filename) shutil.copy("data/empty_audio.ogg", filename)
@@ -2021,13 +2028,14 @@ class KoboldStoryRegister(object):
text = [text] text = [text]
output = None output = None
for process_text in text: for process_text in text:
audio = self.tortoise.tts_with_preset(process_text, preset='fast').numpy() audio = self.tortoise.tts_with_preset(process_text, preset='ultra_fast', voice_samples=voice_samples, conditioning_latents=conditioning_latents).numpy()
channels = 2 if (audio.ndim == 2 and audio.shape[1] == 2) else 1 channels = 2 if (audio.ndim == 2 and audio.shape[1] == 2) else 1
if output is None: if output is None:
output = pydub.AudioSegment(np.int16(audio * 2 ** 15).tobytes(), frame_rate=sample_rate, sample_width=2, channels=channels) output = pydub.AudioSegment(np.int16(audio * 2 ** 15).tobytes(), frame_rate=sample_rate, sample_width=2, channels=channels)
else: else:
output = output + pydub.AudioSegment(np.int16(audio * 2 ** 15).tobytes(), frame_rate=sample_rate, sample_width=2, channels=channels) output = output + pydub.AudioSegment(np.int16(audio * 2 ** 15).tobytes(), frame_rate=sample_rate, sample_width=2, channels=channels)
output.export(filename, format="ogg", bitrate="16k") output.export(filename, format="ogg", bitrate="16k")
logger.info("Slow audio took {} for {} characters".format(time.time()-start_time, text_length))
def gen_all_audio(self, overwrite=False): def gen_all_audio(self, overwrite=False):