Skip to content

SOTA Open Source TTS

update deps & remove audios

Lengyue6e88efb5d8f9
diff --git a/docs/assets/audios/0_input.wav b/docs/assets/audios/0_input.wav
deleted file mode 100644
index 6234b49..0000000
Binary files a/docs/assets/audios/0_input.wav and /dev/null differ
diff --git a/docs/assets/audios/0_output.wav b/docs/assets/audios/0_output.wav
deleted file mode 100644
index 3a8c260..0000000
Binary files a/docs/assets/audios/0_output.wav and /dev/null differ
diff --git a/docs/assets/audios/10_output.wav b/docs/assets/audios/10_output.wav
deleted file mode 100644
index 279020b..0000000
Binary files a/docs/assets/audios/10_output.wav and /dev/null differ
diff --git a/docs/assets/audios/11_output.wav b/docs/assets/audios/11_output.wav
deleted file mode 100644
index e7e97ef..0000000
Binary files a/docs/assets/audios/11_output.wav and /dev/null differ
diff --git a/docs/assets/audios/1_input.wav b/docs/assets/audios/1_input.wav
deleted file mode 100644
index 21771cd..0000000
Binary files a/docs/assets/audios/1_input.wav and /dev/null differ
diff --git a/docs/assets/audios/1_output.wav b/docs/assets/audios/1_output.wav
deleted file mode 100644
index cde0e60..0000000
Binary files a/docs/assets/audios/1_output.wav and /dev/null differ
diff --git a/docs/assets/audios/2_input.wav b/docs/assets/audios/2_input.wav
deleted file mode 100644
index 95b70cd..0000000
Binary files a/docs/assets/audios/2_input.wav and /dev/null differ
diff --git a/docs/assets/audios/2_output.wav b/docs/assets/audios/2_output.wav
deleted file mode 100644
index 4f47b55..0000000
Binary files a/docs/assets/audios/2_output.wav and /dev/null differ
diff --git a/docs/assets/audios/3_output.wav b/docs/assets/audios/3_output.wav
deleted file mode 100644
index fb5e75f..0000000
Binary files a/docs/assets/audios/3_output.wav and /dev/null differ
diff --git a/docs/assets/audios/4_output.wav b/docs/assets/audios/4_output.wav
deleted file mode 100644
index 96ebc8d..0000000
Binary files a/docs/assets/audios/4_output.wav and /dev/null differ
diff --git a/docs/assets/audios/5_output.wav b/docs/assets/audios/5_output.wav
deleted file mode 100644
index 7ff6807..0000000
Binary files a/docs/assets/audios/5_output.wav and /dev/null differ
diff --git a/docs/assets/audios/6_input.wav b/docs/assets/audios/6_input.wav
deleted file mode 100644
index d6d8435..0000000
Binary files a/docs/assets/audios/6_input.wav and /dev/null differ
diff --git a/docs/assets/audios/6_output.wav b/docs/assets/audios/6_output.wav
deleted file mode 100644
index c21c13a..0000000
Binary files a/docs/assets/audios/6_output.wav and /dev/null differ
diff --git a/docs/assets/audios/7_input.wav b/docs/assets/audios/7_input.wav
deleted file mode 100644
index b7258b2..0000000
Binary files a/docs/assets/audios/7_input.wav and /dev/null differ
diff --git a/docs/assets/audios/7_output.wav b/docs/assets/audios/7_output.wav
deleted file mode 100644
index 2fd669d..0000000
Binary files a/docs/assets/audios/7_output.wav and /dev/null differ
diff --git a/docs/assets/audios/8_output.wav b/docs/assets/audios/8_output.wav
deleted file mode 100644
index cc07f36..0000000
Binary files a/docs/assets/audios/8_output.wav and /dev/null differ
diff --git a/docs/assets/audios/9_output.wav b/docs/assets/audios/9_output.wav
deleted file mode 100644
index a153758..0000000
Binary files a/docs/assets/audios/9_output.wav and /dev/null differ
diff --git a/generated_audios/v1-sft/zh/out_4.wav b/generated_audios/v1-sft/zh/out_4.wav
new file mode 100644
index 0000000..37252bb
Binary files /dev/null and b/generated_audios/v1-sft/zh/out_4.wav differ
diff --git a/generated_audios/v1-sft/zh/out_5.wav b/generated_audios/v1-sft/zh/out_5.wav
new file mode 100644
index 0000000..2b7cd61
Binary files /dev/null and b/generated_audios/v1-sft/zh/out_5.wav differ
diff --git a/tools/api.py b/tools/api.py
index bdc79a3..3672a7e 100644
--- a/tools/api.py
+++ b/tools/api.py
@@ -193,10 +193,10 @@ def parse_args():
parser.add_argument(
"--llama-checkpoint-path",
type=str,
- default="checkpoints/text2semantic-medium-v1-2k.pth",
+ default="checkpoints/text2semantic-sft-large-v1-4k.pth",
)
parser.add_argument(
- "--llama-config-name", type=str, default="dual_ar_2_codebook_medium"
+ "--llama-config-name", type=str, default="dual_ar_2_codebook_large"
)
parser.add_argument(
"--vqgan-checkpoint-path",
diff --git a/tools/webui.py b/tools/webui.py
index 03dd9b2..5d3c204 100644
--- a/tools/webui.py
+++ b/tools/webui.py
@@ -2,14 +2,12 @@ import html
import os
import threading
from argparse import ArgumentParser
-from io import BytesIO
from pathlib import Path
import gradio as gr
import librosa
import torch
from loguru import logger
-from torchaudio import functional as AF
from transformers import AutoTokenizer
from tools.llama.generate import launch_thread_safe_queue
@@ -74,7 +72,10 @@ def inference(
speaker,
):
if args.max_gradio_length > 0 and len(text) > args.max_gradio_length:
- return None, "Text is too long, please keep it under 1000 characters."
+ return (
+ None,
+ f"Text is too long, please keep it under {args.max_gradio_length} characters.",
+ )
# Parse reference audio aka prompt
prompt_tokens = None
@@ -266,10 +267,10 @@ def parse_args():
parser.add_argument(
"--llama-checkpoint-path",
type=Path,
- default="checkpoints/text2semantic-medium-v1-2k.pth",
+ default="checkpoints/text2semantic-sft-large-v1-4k.pth",
)
parser.add_argument(
- "--llama-config-name", type=str, default="dual_ar_2_codebook_medium"
+ "--llama-config-name", type=str, default="dual_ar_2_codebook_large"
)
parser.add_argument(
"--vqgan-checkpoint-path",