语音转文本接口，bgm路径修复

2025-10-25 13:06:17 +08:00
parent e4339a5b91
commit d1b4fb6be0
12 changed files with 423 additions and 172 deletions
@@ -0,0 +1,155 @@
+import { execFile } from 'child_process';
+import { promises as fs } from 'fs';
+import os from 'os';
+import path from 'path';
+import { promisify } from 'util';
+
+const execFileAsync = promisify(execFile);
+
+/**
+ * 使用 ffmpeg 从视频二进制中提取第一帧，返回 JPEG buffer
+ */
+export async function extractFirstFrame(videoBuffer: Buffer): Promise<{ buffer: Buffer; contentType: string; ext: string } | null> {
+    const ffmpegCmd = process.env.FFMPEG_PATH || 'ffmpeg';
+    const tmpDir = os.tmpdir();
+    const base = `dy_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`;
+    const inPath = path.join(tmpDir, `${base}.mp4`);
+    const outPath = path.join(tmpDir, `${base}.jpg`);
+
+    try {
+        await fs.writeFile(inPath, videoBuffer);
+        const args = [
+            '-hide_banner',
+            '-loglevel', 'error',
+            '-ss', '0',
+            '-i', inPath,
+            '-frames:v', '1',
+            '-q:v', '2',
+            '-f', 'image2',
+            '-y',
+            outPath,
+        ];
+        await execFileAsync(ffmpegCmd, args, { windowsHide: true });
+        const img = await fs.readFile(outPath);
+        return { buffer: img, contentType: 'image/jpeg', ext: 'jpg' };
+    } catch (e: any) {
+        if (e && (e.code === 'ENOENT' || /not found|is not recognized/i.test(String(e.message)))) {
+            console.warn('系统未检测到 ffmpeg，可安装并配置 PATH 或设置 FFMPEG_PATH 后启用封面提取。');
+            return null;
+        }
+        throw e;
+    } finally {
+        try { await fs.unlink(inPath); } catch { }
+        try { await fs.unlink(outPath); } catch { }
+    }
+}
+
+/**
+ * 使用 ffprobe 获取视频时长（毫秒）
+ */
+export async function getVideoDuration(videoBuffer: Buffer): Promise<number | null> {
+    const ffprobeCmd = process.env.FFPROBE_PATH || 'ffprobe';
+    const tmpDir = os.tmpdir();
+    const base = `dy_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`;
+    const inPath = path.join(tmpDir, `${base}.mp4`);
+
+    try {
+        await fs.writeFile(inPath, videoBuffer);
+        const args = [
+            '-v', 'error',
+            '-show_entries', 'format=duration',
+            '-of', 'default=noprint_wrappers=1:nokey=1',
+            inPath,
+        ];
+        const { stdout } = await execFileAsync(ffprobeCmd, args, { windowsHide: true });
+        const durationSeconds = parseFloat(stdout.trim());
+        if (isNaN(durationSeconds)) return null;
+        return Math.round(durationSeconds * 1000); // 转换为毫秒
+    } catch (e: any) {
+        if (e && (e.code === 'ENOENT' || /not found|is not recognized/i.test(String(e.message)))) {
+            console.warn('系统未检测到 ffprobe，可安装并配置 PATH 或设置 FFPROBE_PATH 后启用时长提取。');
+            return null;
+        }
+        console.warn(`获取视频时长失败: ${e?.message || e}`);
+        return null;
+    } finally {
+        try { await fs.unlink(inPath); } catch { }
+    }
+}
+
+/**
+ * 使用 ffmpeg 将视频二进制转换为音频，返回音频 Buffer
+ *
+ * 默认输出 mp3，可通过 opts.format 指定 'mp3' | 'aac' | 'wav'
+ * - mp3: 使用 libmp3lame，contentType 为 'audio/mpeg'
+ * - aac: 输出 m4a 容器（AAC 编码），contentType 为 'audio/mp4'
+ * - wav: PCM 16-bit，44100Hz，双声道，contentType 为 'audio/wav'
+ *
+ * 若系统未安装 ffmpeg 或未在 PATH 中，返回 null 并给出提示。
+ */
+export async function extractAudio(
+    videoBuffer: Buffer,
+    opts?: { format?: 'mp3' | 'aac' | 'wav'; bitrateKbps?: number }
+): Promise<{ buffer: Buffer; contentType: string; ext: string } | null> {
+    const ffmpegCmd = process.env.FFMPEG_PATH || 'ffmpeg';
+    const tmpDir = os.tmpdir();
+    const base = `dy_${Date.now()}_${Math.random().toString(36).slice(2, 8)}`;
+    const inPath = path.join(tmpDir, `${base}.mp4`);
+
+    const format = opts?.format ?? 'mp3';
+    const bitrate = Math.max(32, Math.min(512, opts?.bitrateKbps ?? 192)); // 安全范围 32~512 kbps
+
+    // 根据目标格式设置输出路径、MIME 与编码参数
+    let outPath = '';
+    let contentType = '';
+    let ext = '';
+    let codecArgs: string[] = [];
+
+    if (format === 'mp3') {
+        ext = 'mp3';
+        contentType = 'audio/mpeg';
+        outPath = path.join(tmpDir, `${base}.${ext}`);
+        codecArgs = ['-c:a', 'libmp3lame', '-b:a', `${bitrate}k`];
+    } else if (format === 'aac') {
+        // 使用 m4a 容器更通用
+        ext = 'm4a';
+        contentType = 'audio/mp4';
+        outPath = path.join(tmpDir, `${base}.${ext}`);
+        codecArgs = ['-c:a', 'aac', '-b:a', `${bitrate}k`, '-movflags', '+faststart'];
+    } else {
+        // wav
+        ext = 'wav';
+        contentType = 'audio/wav';
+        outPath = path.join(tmpDir, `${base}.${ext}`);
+        codecArgs = ['-f', 'wav', '-acodec', 'pcm_s16le', '-ar', '44100', '-ac', '2'];
+    }
+
+    try {
+        await fs.writeFile(inPath, videoBuffer);
+        const args = [
+            '-hide_banner',
+            '-loglevel', 'error',
+            '-i', inPath,
+            '-vn', // 丢弃视频流
+            ...codecArgs,
+            '-y',
+            outPath,
+        ];
+        await execFileAsync(ffmpegCmd, args, { windowsHide: true });
+        const audio = await fs.readFile(outPath);
+        return { buffer: audio, contentType, ext };
+    } catch (e: any) {
+        if (e && (e.code === 'ENOENT' || /not found|is not recognized/i.test(String(e.message)))) {
+            console.warn('系统未检测到 ffmpeg，可安装并配置 PATH 或设置 FFMPEG_PATH 后启用音频提取。');
+            return null;
+        }
+        // 一些环境可能缺少特定编码器（如 libmp3lame），提示并抛出原始错误
+        console.warn(`提取音频失败: ${e?.message || e}`);
+        throw e;
+    } finally {
+        try { await fs.unlink(inPath); } catch { }
+        try {
+            if (outPath) await fs.unlink(outPath);
+        } catch { }
+    }
+}