Hugging Face's logo

Spaces:
KGNINJA
/
MLIT_DATA_PLATFORM
Runtime error

App Files Community
MLIT_DATA_PLATFORM
File size: 2,257 Bytes 
1b9afe7
 
 
 
848980e
1b9afe7
 
848980e
1b9afe7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
 
import io
import numpy as np
import librosa
import pyworld as pw
import gradio as gr
from openai import OpenAI
import soundfile as sf

client = OpenAI()

# ==========================
# 1) 標準の TTS（OpenAI）
# ==========================
def tts_standard(text):
    response = client.audio.speech.create(
        model="gpt-4o-mini-tts",
        voice="alloy",
        input=text,
    )
    audio_bytes = response.read()

    # WAVとして読み込み
    audio, sr = sf.read(io.BytesIO(audio_bytes))
    return sr, audio.astype(np.float64)


# ==========================
# 2) Kansai HL F0 カーブ生成
# ==========================
def kansai_hl_curve(length, f0_src):
    nz = np.where(f0_src > 0)[0]
    if len(nz) < 2:
        return f0_src

    low = np.percentile(f0_src[nz], 20)
    high = np.percentile(f0_src[nz], 85)

    L = length
    p1 = int(L * 0.25)
    p2 = int(L * 0.70)
    p3 = L

    seg1 = np.linspace(low, high, p1)
    seg2 = np.linspace(high, high, p2 - p1)
    seg3 = np.linspace(high, low, p3 - p2)

    out = np.concatenate([seg1, seg2, seg3])

    if len(out) < length:
        out = np.pad(out, (0, length - len(out)), mode="edge")
    else:
        out = out[:length]

    return out


# ==========================
# 3) 標準 → 関西イントネーション化
# ==========================
def convert_to_kansai_pitch(audio, sr):
    f0, sp, ap = pw.wav2world(audio, sr)
    f0_new = kansai_hl_curve(len(f0), f0)
    y = pw.synthesize(f0_new, sp, ap, sr)
    return y.astype(np.float32)


# ==========================
# 4) 統合
# ==========================
def kansai_tts(text):
    sr, audio_std = tts_standard(text)
    audio_ks = convert_to_kansai_pitch(audio_std, sr)
    return (sr, audio_ks)


# ==========================
# 5) Gradio UI
# ==========================
with gr.Blocks() as demo:
    gr.Markdown("## 🎙 Kansign — Kansai Accent TTS（本物HLイントネーション搭載）")

    text_in = gr.Textbox(label="テキストを入力（例：なんでやねん）")
    audio_out = gr.Audio(label="関西イントネーション音声", type="numpy")
    btn = gr.Button("関西イントネーションで喋る")

    btn.click(kansai_tts, inputs=text_in, outputs=audio_out)

demo.launch()