app.py

import base64
import json
import logging
import os
import re
import sys
import tempfile
from datetime import datetime, timezone
from pathlib import Path
from threading import Lock
from typing import Optional, Tuple

import gradio as gr
import numpy as np

os.environ["TOKENIZERS_PARALLELISM"] = "false"
os.environ.setdefault("OPENBLAS_NUM_THREADS", "4")
os.environ.setdefault("OMP_NUM_THREADS", "4")
os.environ.setdefault("MKL_NUM_THREADS", "4")

logging.basicConfig(
    level=logging.INFO,
    format="%(asctime)s - %(levelname)s - %(message)s",
    handlers=[logging.StreamHandler(sys.stdout)],
)
logger = logging.getLogger(__name__)

NANOVLLM_API_BASE = os.environ.get("NANOVLLM_API_BASE", "").rstrip("/")
if not NANOVLLM_API_BASE:
    raise RuntimeError("NANOVLLM_API_BASE environment variable is required but not set.")
DEFAULT_ASR_MODEL_REF = "FunAudioLLM/SenseVoiceSmall"
MAX_REFERENCE_AUDIO_SECONDS = 50.0

_persistent_root = None
_request_log_dir = None
_active_generation_requests = 0
_active_generation_lock = Lock()


def _get_int_env(name: str, default: int) -> int:
    value = os.environ.get(name, "").strip()
    if not value:
        return default
    return int(value)


def _get_float_env(name: str, default: float) -> float:
    value = os.environ.get(name, "").strip()
    if not value:
        return default
    return float(value)


def _get_bool_env(name: str, default: bool) -> bool:
    value = os.environ.get(name, "").strip().lower()
    if not value:
        return default
    if value in {"1", "true", "yes", "on"}:
        return True
    if value in {"0", "false", "no", "off"}:
        return False
    raise ValueError(f"Invalid boolean env: {name}={value!r}")


# ---------- Request Logging ----------


def _configure_cache_dirs() -> None:
    global _persistent_root, _request_log_dir
    persistent_root = Path(os.environ.get("SPACE_PERSISTENT_ROOT", "/data")).expanduser()
    if not persistent_root.exists():
        logger.info("Persistent storage not detected. Request logs disabled.")
        return

    logs_dir = Path(
        os.environ.get("REQUEST_LOG_DIR", str(persistent_root / "logs"))
    ).expanduser()
    logs_dir.mkdir(parents=True, exist_ok=True)
    _persistent_root = persistent_root
    _request_log_dir = logs_dir
    logger.info(f"Persistent storage detected at {persistent_root}")
    logger.info(f"Request logs will be written to daily files under {_request_log_dir}")


_configure_cache_dirs()


def _append_request_log(payload: dict) -> None:
    if _request_log_dir is None:
        return
    now = datetime.now(timezone.utc)
    record = {"timestamp": now.isoformat(), **payload}
    log_path = _request_log_dir / f"{now.date().isoformat()}.jsonl"
    with log_path.open("a", encoding="utf-8") as fp:
        fp.write(json.dumps(record, ensure_ascii=False) + "\n")


def _begin_generation_request() -> None:
    global _active_generation_requests
    with _active_generation_lock:
        _active_generation_requests += 1


def _end_generation_request() -> None:
    global _active_generation_requests
    with _active_generation_lock:
        _active_generation_requests = max(0, _active_generation_requests - 1)


def _get_active_generation_requests() -> int:
    with _active_generation_lock:
        return _active_generation_requests


# ---------- Remote ASR & Denoise via HTTP API ----------


def _api_asr(audio_path: str) -> str:
    """Call POST /asr on the nanovllm server to transcribe audio."""
    import requests

    path = Path(audio_path)
    wav_b64 = base64.b64encode(path.read_bytes()).decode("utf-8")
    wav_fmt = path.suffix.lstrip(".").lower() or "wav"

    resp = requests.post(
        f"{NANOVLLM_API_BASE}/asr",
        json={"wav_base64": wav_b64, "wav_format": wav_fmt},
        timeout=60,
    )
    resp.raise_for_status()
    return resp.json().get("text", "")


def _api_denoise(audio_path: str) -> str:
    """Call POST /denoise on the nanovllm server, return path to denoised temp file."""
    import requests

    path = Path(audio_path)
    wav_b64 = base64.b64encode(path.read_bytes()).decode("utf-8")
    wav_fmt = path.suffix.lstrip(".").lower() or "wav"

    resp = requests.post(
        f"{NANOVLLM_API_BASE}/denoise",
        json={"wav_base64": wav_b64, "wav_format": wav_fmt},
        timeout=120,
    )
    resp.raise_for_status()

    denoised_b64 = resp.json()["wav_base64"]
    denoised_bytes = base64.b64decode(denoised_b64)

    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
    tmp.write(denoised_bytes)
    tmp.close()
    return tmp.name


# ---------- Text Normalization (CPU-only, from VoxCPM text_normalize.py) ----------

_chinese_char_pattern = re.compile(r"[\u4e00-\u9fff]+")
_text_normalizer = None
_text_normalizer_lock = Lock()


def _contains_chinese(text: str) -> bool:
    return bool(_chinese_char_pattern.search(text))


def _replace_corner_mark(text: str) -> str:
    text = text.replace("\u00b2", "\u5e73\u65b9")
    text = text.replace("\u00b3", "\u7acb\u65b9")
    text = text.replace("\u221a", "\u6839\u53f7")
    text = text.replace("\u2248", "\u7ea6\u7b49\u4e8e")
    text = text.replace("<", "\u5c0f\u4e8e")
    return text


def _remove_bracket(text: str) -> str:
    text = text.replace("\uff08", " ").replace("\uff09", " ")
    text = text.replace("\u3010", " ").replace("\u3011", " ")
    text = text.replace("\u2018", "").replace("\u2019", "")
    text = text.replace("\u2014\u2014", " ")
    return text


def _spell_out_number(text: str, inflect_parser) -> str:
    new_text = []
    st = None
    for i, c in enumerate(text):
        if not c.isdigit():
            if st is not None:
                num_str = inflect_parser.number_to_words(text[st:i])
                new_text.append(num_str)
                st = None
            new_text.append(c)
        else:
            if st is None:
                st = i
    if st is not None and st < len(text):
        num_str = inflect_parser.number_to_words(text[st:])
        new_text.append(num_str)
    return "".join(new_text)


def _replace_blank(text: str) -> str:
    out_str = []
    for i, c in enumerate(text):
        if c == " ":
            if (
                i + 1 < len(text) and text[i + 1].isascii() and text[i + 1] != " "
                and i - 1 >= 0 and text[i - 1].isascii() and text[i - 1] != " "
            ):
                out_str.append(c)
        else:
            out_str.append(c)
    return "".join(out_str)


def _clean_markdown(md_text: str) -> str:
    import regex

    md_text = re.sub(r"```.*?```", "", md_text, flags=re.DOTALL)
    md_text = re.sub(r"`[^`]*`", "", md_text)
    md_text = re.sub(r"!\[[^\]]*\]\([^\)]+\)", "", md_text)
    md_text = re.sub(r"\[([^\]]+)\]\([^)]+\)", r"\1", md_text)
    md_text = re.sub(r"^(\s*)-\s+", r"\1", md_text, flags=re.MULTILINE)
    md_text = re.sub(r"<[^>]+>", "", md_text)
    md_text = re.sub(r"^#{1,6}\s*", "", md_text, flags=re.MULTILINE)
    md_text = re.sub(r"\n\s*\n", "\n", md_text)
    md_text = md_text.strip()
    return md_text


def _clean_text(text: str) -> str:
    import regex

    text = _clean_markdown(text)
    text = regex.compile(r"\p{Emoji_Presentation}|\p{Emoji}\uFE0F", flags=regex.UNICODE).sub("", text)
    text = text.replace("\n", " ").replace("\t", " ")
    text = text.replace("\u201c", '"').replace("\u201d", '"')
    return text


def _get_text_normalizer():
    global _text_normalizer
    if _text_normalizer is not None:
        return _text_normalizer
    with _text_normalizer_lock:
        if _text_normalizer is not None:
            return _text_normalizer
        from wetext import Normalizer
        import inflect

        _text_normalizer = {
            "zh_tn": Normalizer(lang="zh", operator="tn", remove_erhua=True),
            "en_tn": Normalizer(lang="en", operator="tn"),
            "inflect": inflect.engine(),
        }
        logger.info("TextNormalizer loaded.")
    return _text_normalizer


def normalize_text(text: str) -> str:
    """Normalize text (numbers, dates, abbreviations) for TTS input."""
    tn = _get_text_normalizer()
    lang = "zh" if _contains_chinese(text) else "en"
    text = _clean_text(text)
    if lang == "zh":
        text = text.replace("=", "\u7b49\u4e8e")
        if re.search(r"([\d$%^*_+\u2265\u2264\u2260\u00d7\u00f7?=])", text):
            text = re.sub(r"(?<=[a-zA-Z0-9])-(?=\d)", " - ", text)
        text = tn["zh_tn"].normalize(text)
        text = _replace_blank(text)
        text = _replace_corner_mark(text)
        text = _remove_bracket(text)
    else:
        text = tn["en_tn"].normalize(text)
        text = _spell_out_number(text, tn["inflect"])
    return text


def _safe_prompt_wav_recognition(
    use_prompt_text: bool, prompt_wav: Optional[str], request: Optional[gr.Request] = None
) -> str:
    if not use_prompt_text or prompt_wav is None or not prompt_wav.strip():
        return ""
    try:
        return _api_asr(prompt_wav)
    except Exception as exc:
        logger.warning(f"ASR recognition failed: {exc}")
        raise gr.Error(_get_i18n_text("asr_failed_error", request)) from exc


# ---------- Audio helpers ----------


def _get_audio_duration_seconds(audio_path: str) -> float:
    import soundfile as sf
    try:
        info = sf.info(audio_path)
        return float(info.frames) / float(info.samplerate)
    except Exception:
        pass
    import subprocess
    try:
        result = subprocess.run(
            ["ffprobe", "-v", "error", "-show_entries", "format=duration",
             "-of", "default=noprint_wrappers=1:nokey=1", audio_path],
            capture_output=True, text=True, timeout=10,
        )
        if result.returncode == 0 and result.stdout.strip():
            return float(result.stdout.strip())
    except (FileNotFoundError, subprocess.TimeoutExpired, ValueError):
        pass
    raise RuntimeError(f"Cannot determine audio duration: {audio_path}")


def _validate_reference_audio_duration(
    audio_path: str, request: Optional[gr.Request] = None
) -> None:
    duration_seconds = _get_audio_duration_seconds(audio_path)
    if duration_seconds > MAX_REFERENCE_AUDIO_SECONDS:
        raise gr.Error(_get_i18n_text("reference_audio_too_long_error", request))


# ---------- Nano-vLLM HTTP API Client ----------


def _api_generate(payload: dict) -> str:
    """Call POST /generate, receive streaming MP3, save to temp file and return path."""
    import requests

    url = f"{NANOVLLM_API_BASE}/generate"
    logger.info(f"Calling {url} ...")

    resp = requests.post(url, json=payload, stream=True, timeout=300)
    resp.raise_for_status()

    tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
    try:
        for chunk in resp.iter_content(chunk_size=64 * 1024):
            tmp.write(chunk)
        tmp.close()
        return tmp.name
    except Exception:
        tmp.close()
        if os.path.exists(tmp.name):
            os.unlink(tmp.name)
        raise


def _api_get_info() -> dict:
    import requests

    resp = requests.get(f"{NANOVLLM_API_BASE}/info", timeout=10)
    resp.raise_for_status()
    return resp.json()


# ---------- Generation via HTTP API ----------


def generate_tts_audio(
    text_input: str,
    control_instruction: str = "",
    reference_wav_path_input: Optional[str] = None,
    use_prompt_text: bool = False,
    prompt_text_input: str = "",
    cfg_value_input: float = 2.0,
    do_normalize: bool = True,
    denoise: bool = True,
    request: Optional[gr.Request] = None,
) -> str:
    _begin_generation_request()
    request_payload = {
        "event": "tts_request",
        "ui_language": _resolve_ui_language(request),
        "text": (text_input or "").strip(),
        "control_instruction": (control_instruction or "").strip(),
        "use_prompt_text": bool(use_prompt_text),
        "prompt_text": (prompt_text_input or "").strip(),
        "cfg_value": float(cfg_value_input),
        "do_normalize": bool(do_normalize),
        "denoise": bool(denoise),
        "has_reference_audio": bool(reference_wav_path_input and reference_wav_path_input.strip()),
    }
    if request_payload["has_reference_audio"]:
        try:
            request_payload["reference_audio_duration_seconds"] = round(
                _get_audio_duration_seconds(reference_wav_path_input), 3
            )
        except Exception as exc:
            request_payload["reference_audio_duration_error"] = str(exc)

    try:
        text = (text_input or "").strip()
        if not text:
            raise ValueError("Please input text to synthesize.")

        control = (control_instruction or "").strip()
        final_text = f"({control}){text}" if control and not use_prompt_text else text

        if do_normalize:
            try:
                original = final_text
                final_text = normalize_text(final_text)
                if final_text != original:
                    logger.info(f"Text normalized: '{original[:60]}' -> '{final_text[:60]}'")
            except Exception as exc:
                logger.warning(f"Text normalization failed, using original: {exc}")

        prompt_text_clean = (prompt_text_input or "").strip()
        if use_prompt_text and not reference_wav_path_input:
            raise ValueError("Ultimate Cloning Mode requires a reference audio clip.")
        if use_prompt_text and not prompt_text_clean:
            raise ValueError(
                "Ultimate Cloning Mode requires a transcript. "
                "Please wait for ASR or fill it in manually."
            )
        if not use_prompt_text:
            prompt_text_clean = ""

        has_ref = reference_wav_path_input and reference_wav_path_input.strip()
        if has_ref:
            _validate_reference_audio_duration(reference_wav_path_input, request)

        denoised_tmp = None
        api_payload: dict = {
            "target_text": final_text,
            "cfg_value": float(cfg_value_input),
        }

        try:
            if has_ref:
                actual_ref_path = reference_wav_path_input
                if denoise:
                    logger.info("Applying server-side denoise to reference audio ...")
                    try:
                        denoised_tmp = _api_denoise(reference_wav_path_input)
                        actual_ref_path = denoised_tmp
                        logger.info("Denoise completed.")
                    except Exception as exc:
                        logger.warning(f"Denoise failed, using original audio: {exc}")

                ref_path = Path(actual_ref_path)
                wav_b64 = base64.b64encode(ref_path.read_bytes()).decode("utf-8")
                wav_fmt = ref_path.suffix.lstrip(".").lower() or "wav"

                if use_prompt_text:
                    logger.info("[Ultimate Cloning] reference audio + transcript")
                    api_payload["prompt_wav_base64"] = wav_b64
                    api_payload["prompt_wav_format"] = wav_fmt
                    api_payload["prompt_text"] = prompt_text_clean
                    api_payload["ref_audio_wav_base64"] = wav_b64
                    api_payload["ref_audio_wav_format"] = wav_fmt
                else:
                    logger.info("[Controllable Cloning] reference audio only")
                    api_payload["ref_audio_wav_base64"] = wav_b64
                    api_payload["ref_audio_wav_format"] = wav_fmt
            else:
                logger.info(f"[Voice Design] control: {control[:50] if control else 'None'}")

            logger.info(f"Generating: '{final_text[:80]}...'")
            mp3_path = _api_generate(api_payload)
        finally:
            if denoised_tmp and os.path.exists(denoised_tmp):
                try:
                    os.unlink(denoised_tmp)
                except OSError:
                    pass

        try:
            _append_request_log({**request_payload, "status": "success"})
        except Exception as exc:
            logger.warning(f"Failed to append request log: {exc}")

        return mp3_path

    except (ValueError, gr.Error) as exc:
        try:
            _append_request_log({**request_payload, "status": "rejected", "error": str(exc)})
        except Exception:
            pass
        if isinstance(exc, gr.Error):
            raise
        raise gr.Error(str(exc)) from exc
    except Exception as exc:
        logger.exception("Generation failed")
        try:
            _append_request_log({**request_payload, "status": "error", "error": str(exc)})
        except Exception:
            pass
        raise gr.Error(_get_i18n_text("backend_retry_error", request)) from exc
    finally:
        _end_generation_request()


# ---------- Inline i18n (en + zh-CN) ----------


_USAGE_INSTRUCTIONS_EN = (
    "**VoxCPM2 — Three Modes of Speech Generation:**\n\n"
    "🎨 **Voice Design** — Create a brand-new voice  \n"
    "No reference audio required. Describe the desired voice characteristics "
    "(gender, age, tone, emotion, pace …) in **Control Instruction**, and VoxCPM2 "
    "will craft a unique voice from your description alone.\n\n"
    "🎛️ **Controllable Cloning** — Clone a voice with optional style guidance  \n"
    "Upload a reference audio clip, then use **Control Instruction** to steer "
    "emotion, speaking pace, and overall style while preserving the original timbre.\n\n"
    "🎙️ **Ultimate Cloning** — Reproduce every vocal nuance through audio continuation  \n"
    "Turn on **Ultimate Cloning Mode** and provide (or auto-transcribe) the reference audio's transcript. "
    "The model treats the reference clip as a spoken prefix and seamlessly **continues** from it, faithfully preserving every vocal detail."
    "Note: This mode will disable Control Instruction.\n\n"
    "### [A Voice Chef's Guide to VoxCPM2](https://voxcpm.readthedocs.io/en/latest/cookbook.html)"
)

_EXAMPLES_FOOTER_EN = (
    "---\n"
    "**💡 Voice Description Examples:**  \n"
    "Try the following Control Instructions to explore different voices:  \n\n"
    "**Example 1 — Gentle & Melancholic Girl**  \n"
    '`Control Instruction`: *"A young girl with a soft, sweet voice. '
    'Speaks slowly with a melancholic, slightly tsundere tone."*  \n'
    '`Target Text`: *"I never asked you to stay… It\'s not like I care or anything. '
    'But… why does it still hurt so much now that you\'re gone?"*  \n\n'
    "**Example 2 — Laid-Back Surfer Dude**  \n"
    '`Control Instruction`: *"Relaxed young male voice, slightly nasal, '
    'lazy drawl, very casual and chill."*  \n'
    '`Target Text`: *"Dude, did you see that set? The waves out there are totally gnarly today. '
    "Just catching barrels all morning — it's like, totally righteous, you know what I mean?\"*"
)

_USAGE_INSTRUCTIONS_ZH = (
    "**VoxCPM2 — 三种语音生成方式：**\n\n"
    "🎨 **声音设计（Voice Design）**  \n"
    "无需参考音频。在 **Control Instruction** 中描述目标音色特征"
    "（性别、年龄、语气、情绪、语速等），VoxCPM2 即可为你从零创造独一无二的声音。\n\n"
    "🎛️ **可控克隆（Controllable Cloning）**  \n"
    "上传参考音频，同时可选地使用 **Control Instruction** 来指定情绪、语速、风格等表达方式，"
    "在保留原始音色的基础上灵活控制说话风格。\n\n"
    "🎙️ **极致克隆（Ultimate Cloning）**  \n"
    "开启 **极致克隆模式** 并提供参考音频的文字内容（可自动识别）。"
    "模型会将参考音频视为已说出的前文，以**音频续写**的方式完整还原参考音频中的所有声音细节。"
    "注意：该模式与可控克隆模式互斥，将禁用Control Instruction。\n\n"
    "### [VoxCPM 2 最佳实践指南](https://voxcpm.readthedocs.io/en/latest/cookbook.html)\n\n"
)

_EXAMPLES_FOOTER_ZH = (
    "---\n"
    "**💡 声音描述示例（中英文均可）：**  \n\n"
    "**示例 1 — 深宫太后**  \n"
    '`Control Instruction`: *"中老年女性，声音低沉阴冷，语速缓慢而有力，'
    '字字深思熟虑，带有深不可测的城府与威慑感。"*  \n'
    '`Target Text`: *"哀家在这深宫待了四十年，什么风浪没见过？你以为瞒得过哀家？"*  \n\n'
    "**示例 2 — 暴躁驾校教练**  \n"
    '`Control Instruction`: *"暴躁的中年男声，语速快，充满无奈和愤怒"*  \n'
    '`Target Text`: *"踩离合！踩刹车啊！你往哪儿开呢？前面是树你看不见吗？'
    '我教了你八百遍了，打死方向盘！你是不是想把车给我开到沟里去？"*  \n\n'
    "---\n"
    "**🗣️ 方言生成指南：**  \n"
    "要生成地道的方言语音，请在 **Target Text** 中直接使用方言词汇和句式，"
    "并在 **Control Instruction** 中描述方言特征。  \n\n"
    "**示例 — 广东话**  \n"
    '`Control Instruction`: *"粤语，中年男性，语气平淡"*  \n'
    '✅ 正确（粤语表达）：*"伙計，唔該一個A餐，凍奶茶少甜！"*  \n'
    '❌ 错误（普通话原文）：*"伙计，麻烦来一个A餐，冻奶茶少甜！"*  \n\n'
    "**示例 — 河南话**  \n"
    '`Control Instruction`: *"河南话，接地气的大叔"*  \n'
    '✅ 正确（河南话表达）：*"恁这是弄啥嘞？晌午吃啥饭？"*  \n'
    '❌ 错误（普通话原文）：*"你这是在干什么呢？中午吃什么饭？"*  \n\n'
    "🤖 **小技巧：** 不知道方言怎么写？可以用豆包、DeepSeek、Kimi 等 AI 助手"
    "将普通话翻译为方言文本，再粘贴到 Target Text 中即可。  \n\n"
)

_I18N_TRANSLATIONS = {
    "en": {
        "reference_audio_label": "🎤 Reference Audio (optional — upload for cloning)",
        "show_prompt_text_label": "🎙️ Ultimate Cloning Mode (transcript-guided cloning)",
        "show_prompt_text_info": "Auto-transcribes reference audio for every vocal nuance reproduced. Control Instruction will be disabled when active.",
        "prompt_text_label": "Transcript of Reference Audio (auto-filled via ASR, editable)",
        "prompt_text_placeholder": "The transcript of your reference audio will appear here …",
        "control_label": "🎛️ Control Instruction (optional — supports Chinese & English)",
        "control_placeholder": "e.g. A warm young woman / 年轻女性，温柔甜美 / Excited and fast-paced",
        "target_text_label": "✍️ Target Text — the content to speak",
        "generate_btn": "🔊 Generate Speech",
        "generated_audio_label": "Generated Audio",
        "advanced_settings_title": "⚙️ Advanced Settings",
        "ref_denoise_label": "Reference audio enhancement",
        "ref_denoise_info": "Apply ZipEnhancer denoising to the reference audio before cloning",
        "normalize_label": "Text normalization",
        "normalize_info": "Normalize numbers, dates, and abbreviations via wetext",
        "cfg_label": "CFG (guidance scale)",
        "cfg_info": "Higher → closer to the prompt / reference; lower → more creative variation",
        "reference_audio_too_long_error": "Reference audio is too long. Please upload audio no longer than 50 seconds.",
        "denoise_busy_error": "Too many reference-audio enhancement requests are running. Please try again in a moment.",
        "denoise_failed_error": "Reference audio enhancement failed. Please try disabling denoise or use a cleaner clip.",
        "backend_retry_error": "The backend is temporarily unstable. Please try again in a moment.",
        "asr_failed_error": "ASR failed. Please fill the transcript manually or try another reference audio.",
        "usage_instructions": _USAGE_INSTRUCTIONS_EN,
        "examples_footer": _EXAMPLES_FOOTER_EN,
    },
    "zh-CN": {
        "reference_audio_label": "🎤 参考音频（可选 — 上传后用于克隆）",
        "show_prompt_text_label": "🎙️ 极致克隆模式（基于文本引导的极致克隆）",
        "show_prompt_text_info": "自动识别参考音频文本，完整还原音色、节奏、情感等全部声音细节。开启后 Control Instruction 将暂时禁用",
        "prompt_text_label": "参考音频内容文本（ASR 自动填充，可手动编辑）",
        "prompt_text_placeholder": "参考音频的文字内容将自动识别并显示在此处 …",
        "control_label": "🎛️ Control Instruction（可选 — 支持中英文描述）",
        "control_placeholder": "如：年轻女性，温柔甜美 / A warm young woman / 暴躁老哥，语速飞快",
        "target_text_label": "✍️ Target Text — 要合成的目标文本",
        "generate_btn": "🔊 开始生成",
        "generated_audio_label": "生成结果",
        "advanced_settings_title": "⚙️ 高级设置",
        "ref_denoise_label": "参考音频降噪增强",
        "ref_denoise_info": "克隆前使用 ZipEnhancer 对参考音频进行降噪处理",
        "normalize_label": "文本规范化",
        "normalize_info": "自动规范化数字、日期及缩写（基于 wetext）",
        "cfg_label": "CFG（引导强度）",
        "cfg_info": "数值越高 → 越贴合提示/参考音色；数值越低 → 生成风格更自由",
        "reference_audio_too_long_error": "参考音频太长了，请上传不超过 50 秒的音频。",
        "denoise_busy_error": "当前参考音频降噪请求过多，请稍后再试。",
        "denoise_failed_error": "参考音频降噪失败，请尝试关闭降噪或更换更干净的音频。",
        "backend_retry_error": "后端暂时不稳定，请稍后再试。",
        "asr_failed_error": "ASR 识别失败，请手动填写参考音频文本，或更换一段参考音频后重试。",
        "usage_instructions": _USAGE_INSTRUCTIONS_ZH,
        "examples_footer": _EXAMPLES_FOOTER_ZH,
    },
    "zh-Hans": None,
    "zh": None,
}
_I18N_TRANSLATIONS["zh-Hans"] = _I18N_TRANSLATIONS["zh-CN"]
_I18N_TRANSLATIONS["zh"] = _I18N_TRANSLATIONS["zh-CN"]

for _d in _I18N_TRANSLATIONS.values():
    if _d is not None:
        for _k, _v in _I18N_TRANSLATIONS["en"].items():
            _d.setdefault(_k, _v)

I18N = gr.I18n(**_I18N_TRANSLATIONS)


def _resolve_ui_language(request: Optional[gr.Request] = None) -> str:
    if request is None:
        return "en"
    accept_language = str(request.headers.get("accept-language", "")).lower()
    if accept_language.startswith("zh"):
        return "zh-CN"
    return "en"


def _get_i18n_text(key: str, request: Optional[gr.Request] = None) -> str:
    locale = _resolve_ui_language(request)
    return _I18N_TRANSLATIONS.get(locale, _I18N_TRANSLATIONS["en"]).get(
        key, _I18N_TRANSLATIONS["en"].get(key, key)
    )


# ---------- Theme & CSS ----------

DEFAULT_TARGET_TEXT = (
    "VoxCPM2 is a creative multilingual TTS model from ModelBest, "
    "designed to generate highly realistic speech."
)

_CUSTOM_CSS = """
.logo-container {
    text-align: center;
    margin: 0.5rem 0 1rem 0;
}
.logo-container img {
    height: 80px;
    width: auto;
    max-width: 200px;
    display: inline-block;
}

/* Toggle switch style */
.switch-toggle {
    padding: 8px 12px;
    border-radius: 8px;
    background: var(--block-background-fill);
}
.switch-toggle input[type="checkbox"] {
    appearance: none;
    -webkit-appearance: none;
    width: 44px;
    height: 24px;
    background: #ccc;
    border-radius: 12px;
    position: relative;
    cursor: pointer;
    transition: background 0.3s ease;
    flex-shrink: 0;
}
.switch-toggle input[type="checkbox"]::after {
    content: "";
    position: absolute;
    top: 2px;
    left: 2px;
    width: 20px;
    height: 20px;
    background: white;
    border-radius: 50%;
    transition: transform 0.3s ease;
    box-shadow: 0 1px 3px rgba(0,0,0,0.2);
}
.switch-toggle input[type="checkbox"]:checked {
    background: var(--color-accent);
}
.switch-toggle input[type="checkbox"]:checked::after {
    transform: translateX(20px);
}
"""

_APP_THEME = gr.themes.Soft(
    primary_hue="blue",
    secondary_hue="gray",
    neutral_hue="slate",
    font=[gr.themes.GoogleFont("Inter"), "Arial", "sans-serif"],
)


# ---------- UI ----------


def create_demo_interface():
    assets_dir = Path.cwd().absolute() / "assets"
    if assets_dir.exists():
        gr.set_static_paths(paths=[assets_dir])

    def _on_toggle_instant(checked):
        if checked:
            return (
                gr.update(visible=True, value="", placeholder="Recognizing reference audio..."),
                gr.update(visible=False),
            )
        return (
            gr.update(visible=False),
            gr.update(visible=True, interactive=True),
        )

    def _run_asr_if_needed(checked, audio_path, request: gr.Request = None):
        if not checked or not audio_path:
            return gr.update()
        logger.info("Running ASR on reference audio...")
        asr_text = _safe_prompt_wav_recognition(True, audio_path, request=request)
        logger.info(f"ASR result: {asr_text[:60]}...")
        return gr.update(
            value=asr_text,
            placeholder=_get_i18n_text("prompt_text_placeholder", request),
        )

    with gr.Blocks() as interface:
        if (assets_dir / "voxcpm_logo.png").exists():
            gr.HTML(
                '<div class="logo-container">'
                '<img src="/gradio_api/file=assets/voxcpm_logo.png" alt="VoxCPM Logo">'
                "</div>"
            )

        gr.Markdown(I18N("usage_instructions"))

        with gr.Row():
            with gr.Column():
                reference_wav = gr.Audio(
                    sources=["upload", "microphone"],
                    type="filepath",
                    label=I18N("reference_audio_label"),
                )
                show_prompt_text = gr.Checkbox(
                    value=False,
                    label=I18N("show_prompt_text_label"),
                    info=I18N("show_prompt_text_info"),
                    elem_classes=["switch-toggle"],
                )
                prompt_text = gr.Textbox(
                    value="",
                    label=I18N("prompt_text_label"),
                    placeholder=I18N("prompt_text_placeholder"),
                    lines=2,
                    visible=False,
                )
                control_instruction = gr.Textbox(
                    value="",
                    label=I18N("control_label"),
                    placeholder=I18N("control_placeholder"),
                    lines=2,
                )
                text = gr.Textbox(
                    value=DEFAULT_TARGET_TEXT,
                    label=I18N("target_text_label"),
                    lines=3,
                )

                with gr.Accordion(I18N("advanced_settings_title"), open=False):
                    DoDenoisePromptAudio = gr.Checkbox(
                        value=False,
                        label=I18N("ref_denoise_label"),
                        elem_classes=["switch-toggle"],
                        info=I18N("ref_denoise_info"),
                    )
                    DoNormalizeText = gr.Checkbox(
                        value=False,
                        label=I18N("normalize_label"),
                        elem_classes=["switch-toggle"],
                        info=I18N("normalize_info"),
                    )
                    cfg_value = gr.Slider(
                        minimum=1.0,
                        maximum=3.0,
                        value=2.0,
                        step=0.1,
                        label=I18N("cfg_label"),
                        info=I18N("cfg_info"),
                    )
                run_btn = gr.Button(I18N("generate_btn"), variant="primary", size="lg")

            with gr.Column():
                audio_output = gr.Audio(label=I18N("generated_audio_label"))
                gr.Markdown(I18N("examples_footer"))

        show_prompt_text.change(
            fn=_on_toggle_instant,
            inputs=[show_prompt_text],
            outputs=[prompt_text, control_instruction],
        ).then(
            fn=_run_asr_if_needed,
            inputs=[show_prompt_text, reference_wav],
            outputs=[prompt_text],
        )

        run_btn.click(
            fn=generate_tts_audio,
            inputs=[
                text,
                control_instruction,
                reference_wav,
                show_prompt_text,
                prompt_text,
                cfg_value,
                DoNormalizeText,
                DoDenoisePromptAudio,
            ],
            outputs=[audio_output],
            show_progress=True,
            api_name="generate",
        )

    return interface


def run_demo(
    server_name: str = "0.0.0.0", server_port: int = 7860, show_error: bool = True
):
    interface = create_demo_interface()
    interface.queue(
        max_size=_get_int_env("GRADIO_QUEUE_MAX_SIZE", 10),
        default_concurrency_limit=_get_int_env("GRADIO_DEFAULT_CONCURRENCY_LIMIT", 4),
    ).launch(
        server_name=server_name,
        server_port=int(os.environ.get("PORT", server_port)),
        show_error=show_error,
        i18n=I18N,
        theme=_APP_THEME,
        css=_CUSTOM_CSS,
        ssr_mode=_get_bool_env("GRADIO_SSR_MODE", False),
    )


if __name__ == "__main__":
    run_demo()