meijiaka-zy/python-api/app/api/v1/caption.py

"""
火山引擎音视频字幕 API 路由
============================

提供字幕生成、自动打轴等功能。
"""

import logging

from fastapi import APIRouter, HTTPException

from app.schemas.caption import (
    AutoAlignResult,
    AutoAlignSubmitRequest,
    CaptionResult,
    CaptionSubmitRequest,
    CaptionTaskResponse,
    SrtSubtitleResponse,
)
from app.schemas.common import ApiResponse, success_response
from app.services.volcengine_caption_service import (
    VolcengineCaptionError,
    VolcengineCaptionService,
    get_caption_service,
)

logger = logging.getLogger(__name__)
router = APIRouter(prefix="/caption", tags=["Caption"])


@router.post("/submit", response_model=ApiResponse[CaptionTaskResponse])
async def submit_caption_task(request: CaptionSubmitRequest):
    """
    提交字幕生成任务

    提交音频/视频文件URL，生成带时间轴的字幕。
    """
    try:
        service = await get_caption_service()
        task_id = await service.submit_caption_task(
            audio_url=request.audio_url,
            language=request.language,
            caption_type=request.caption_type,
            use_punc=request.use_punc,
            use_itn=request.use_itn,
            words_per_line=request.words_per_line,
            max_lines=request.max_lines,
        )

        return success_response(
            data=CaptionTaskResponse(
                task_id=task_id,
                status="pending",
            ),
            message="字幕任务已提交",
        )

    except VolcengineCaptionError as e:
        logger.error(f"提交字幕任务失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"提交字幕任务异常: {e}")
        raise HTTPException(status_code=500, detail=f"提交失败: {str(e)}")


@router.get("/query/{task_id}", response_model=ApiResponse[CaptionResult])
async def query_caption_task(task_id: str, blocking: bool = True):
    """
    查询字幕任务结果

    Args:
        task_id: 任务ID
        blocking: 是否阻塞等待结果 (默认True)
    """
    try:
        service = await get_caption_service()
        result = await service.query_caption_task(task_id, blocking=blocking)

        return success_response(data=result)

    except VolcengineCaptionError as e:
        logger.error(f"查询字幕任务失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"查询字幕任务异常: {e}")
        raise HTTPException(status_code=500, detail=f"查询失败: {str(e)}")


@router.post("/generate", response_model=ApiResponse[CaptionResult])
async def generate_caption(request: CaptionSubmitRequest, max_wait_time: int = 120):
    """
    生成字幕（完整流程）

    提交任务并轮询结果，直接返回最终字幕数据。
    适用于不需要异步处理的场景。
    """
    try:
        service = await get_caption_service()
        result = await service.generate_caption(
            audio_url=request.audio_url,
            language=request.language,
            caption_type=request.caption_type,
            use_punc=request.use_punc,
            use_itn=request.use_itn,
            words_per_line=request.words_per_line,
            max_lines=request.max_lines,
            max_wait_time=max_wait_time,
        )

        return success_response(data=result)

    except VolcengineCaptionError as e:
        logger.error(f"生成字幕失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"生成字幕异常: {e}")
        raise HTTPException(status_code=500, detail=f"生成失败: {str(e)}")


@router.post("/generate-ass", response_model=ApiResponse[dict])
async def generate_ass(
    request: CaptionSubmitRequest,
    video_width: int = 1080,
    video_height: int = 1920,
    max_wait_time: int = 120,
):
    """
    生成 ASS 格式字幕（完整流程，使用抖音美好体）

    Args:
        video_width: 视频宽度（默认 1080）
        video_height: 视频高度（默认 1920）
    """
    try:
        service = await get_caption_service()
        result = await service.generate_caption(
            audio_url=request.audio_url,
            language=request.language,
            caption_type=request.caption_type,
            use_punc=request.use_punc,
            use_itn=request.use_itn,
            words_per_line=request.words_per_line,
            max_lines=request.max_lines,
            max_wait_time=max_wait_time,
        )

        ass_content = service.to_ass(
            result.utterances,
            video_width=video_width,
            video_height=video_height,
        )

        return success_response(
            data={
                "ass_content": ass_content,
                "utterances": result.utterances,
                "duration": result.duration,
                "font": "DouyinSansBold",
            }
        )

    except Exception as e:
        logger.error(f"生成ASS字幕失败: {e}")
        raise HTTPException(status_code=500, detail=f"生成失败: {str(e)}")


@router.post("/generate-srt", response_model=ApiResponse[SrtSubtitleResponse])
async def generate_srt(request: CaptionSubmitRequest, max_wait_time: int = 120):
    """
    生成 SRT 格式字幕（完整流程）

    直接返回 SRT 格式字幕文件内容。
    """
    try:
        service = await get_caption_service()
        result = await service.generate_caption(
            audio_url=request.audio_url,
            language=request.language,
            caption_type=request.caption_type,
            use_punc=request.use_punc,
            use_itn=request.use_itn,
            words_per_line=request.words_per_line,
            max_lines=request.max_lines,
            max_wait_time=max_wait_time,
        )

        srt_content = service.to_srt(result.utterances)

        return success_response(
            data=SrtSubtitleResponse(
                srt_content=srt_content,
                utterances=result.utterances,
            )
        )

    except VolcengineCaptionError as e:
        logger.error(f"生成SRT字幕失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"生成SRT字幕异常: {e}")
        raise HTTPException(status_code=500, detail=f"生成失败: {str(e)}")


@router.post("/ata/submit", response_model=ApiResponse[CaptionTaskResponse])
async def submit_auto_align_task(request: AutoAlignSubmitRequest):
    """
    提交自动字幕打轴任务

    为已有字幕文本自动配上时间轴。
    """
    try:
        service = await get_caption_service()
        task_id = await service.submit_auto_align_task(
            audio_url=request.audio_url,
            audio_text=request.audio_text,
            caption_type=request.caption_type,
            sta_punc_mode=request.sta_punc_mode,
        )

        return success_response(
            data=CaptionTaskResponse(
                task_id=task_id,
                status="pending",
            ),
            message="打轴任务已提交",
        )

    except VolcengineCaptionError as e:
        logger.error(f"提交打轴任务失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"提交打轴任务异常: {e}")
        raise HTTPException(status_code=500, detail=f"提交失败: {str(e)}")


@router.get("/ata/query/{task_id}", response_model=ApiResponse[AutoAlignResult])
async def query_auto_align_task(task_id: str, blocking: bool = True):
    """
    查询打轴任务结果
    """
    try:
        service = await get_caption_service()
        result = await service.query_auto_align_task(task_id, blocking=blocking)

        return success_response(data=result)

    except VolcengineCaptionError as e:
        logger.error(f"查询打轴任务失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"查询打轴任务异常: {e}")
        raise HTTPException(status_code=500, detail=f"查询失败: {str(e)}")


@router.post("/ata/align")
async def auto_align_caption(request: AutoAlignSubmitRequest, max_wait_time: int = 120):
    """
    自动字幕打轴（完整流程）

    提交打轴任务并轮询结果，直接返回最终数据。
    """
    try:
        logger.info(f"[Caption API] Auto align request: audio_url={request.audio_url[:50]}...")
        service = await get_caption_service()
        result = await service.auto_align_caption(
            audio_url=request.audio_url,
            audio_text=request.audio_text,
            caption_type=request.caption_type,
            sta_punc_mode=request.sta_punc_mode,
            max_wait_time=max_wait_time,
        )
        logger.info(
            f"[Caption API] Auto align result: utterances_count={len(result.utterances) if result.utterances else 0}"
        )
        if result.utterances:
            logger.info(f"[Caption API] First utterance: {result.utterances[0]}")

        # 手动序列化为字典，确保嵌套模型正确处理
        response_data = {
            "code": 0,
            "message": "Success",
            "duration": result.duration,
            "utterances": [
                {
                    "text": u.text,
                    "start_time": u.start_time,
                    "end_time": u.end_time,
                }
                for u in (result.utterances or [])
            ],
        }
        logger.info(f"[Caption API] Response data: {response_data}")
        return success_response(data=response_data)

    except VolcengineCaptionError as e:
        logger.error(f"自动打轴失败: {e}")
        raise HTTPException(status_code=400, detail=str(e))
    except Exception as e:
        logger.error(f"自动打轴异常: {e}")
        raise HTTPException(status_code=500, detail=f"打轴失败: {str(e)}")


@router.post("/convert/ass", response_model=ApiResponse[dict])
async def convert_to_ass(
    result: CaptionResult,
    video_width: int = 1080,
    video_height: int = 1920,
):
    """
    将字幕结果转换为 ASS 格式（使用抖音美好体）
    """
    try:
        service = VolcengineCaptionService("", "")  # 不需要认证
        ass_content = service.to_ass(
            result.utterances,
            video_width=video_width,
            video_height=video_height,
        )

        return success_response(
            data={
                "ass_content": ass_content,
                "font": "DouyinSansBold",
                "utterances_count": len(result.utterances),
            }
        )

    except Exception as e:
        logger.error(f"转换ASS失败: {e}")
        raise HTTPException(status_code=500, detail=f"转换失败: {str(e)}")


@router.post("/convert/srt", response_model=ApiResponse[dict])
async def convert_to_srt(result: CaptionResult):
    """
    将字幕结果转换为 SRT 格式

    用于将 /generate 返回的原始数据转换为 SRT 格式。
    """
    try:
        service = VolcengineCaptionService("", "")  # 不需要认证
        srt_content = service.to_srt(result.utterances)

        return success_response(
            data={
                "srt_content": srt_content,
                "utterances_count": len(result.utterances),
            }
        )

    except Exception as e:
        logger.error(f"转换SRT失败: {e}")
        raise HTTPException(status_code=500, detail=f"转换失败: {str(e)}")


@router.post("/convert/vtt", response_model=ApiResponse[dict])
async def convert_to_vtt(result: CaptionResult):
    """
    将字幕结果转换为 WebVTT 格式
    """
    try:
        service = VolcengineCaptionService("", "")  # 不需要认证
        vtt_content = service.to_vtt(result.utterances)

        return success_response(
            data={
                "vtt_content": vtt_content,
                "utterances_count": len(result.utterances),
            }
        )

    except Exception as e:
        logger.error(f"转换VTT失败: {e}")
        raise HTTPException(status_code=500, detail=f"转换失败: {str(e)}")