vision/preview.go

164 lines
5.2 KiB
Go
Raw Normal View History

package vision
import (
"fmt"
"os"
"os/exec"
"path/filepath"
"strings"
)
// GenerateImagePreview 生成图片预览
// 支持缩放并裁剪以填充指定尺寸 (Fill 模式)
// 根据 outPath 后缀自动转换格式 (.webp, .jpg, .png 等)
func GenerateImagePreview(srcPath, outPath string, width, height int) error {
// 使用统一的 Load() 加载,内部已处理好 HEIC/sips/FFmpeg 的复杂格式兼容
c, err := Load(srcPath)
if err != nil {
return err
}
c.Fill(width, height)
ext := strings.ToLower(filepath.Ext(outPath))
if ext == ".webp" {
// 借用 FFmpeg 将生成的画布转为高质量 WebP (比标准库或第三方库压缩更好)
tmpFile := filepath.Join(os.TempDir(), fmt.Sprintf("preview_%d.png", os.Getpid()))
defer os.Remove(tmpFile)
if err := Save(c, tmpFile); err != nil {
return err
}
v, err := NewVideo()
if err == nil {
cmd := exec.Command(v.FFmpegPath, "-i", tmpFile, "-c:v", "libwebp", "-quality", "80", "-y", outPath)
if err := cmd.Run(); err == nil {
return nil
}
}
}
return Save(c, outPath)
}
// GenerateVideoPreview 生成视频预览
// 根据 outPath 后缀判断输出格式:
// - .webp | .gif: 生成动态动画 (默认每 30 秒采样一帧,可通过 frameInterval 调整)
// - .jpg | .jpeg | .png: 生成单张预览图 (取视频中间帧)
// - 其他: 将 outPath 视为文件夹,在其中生成多张静态 .webp 图像
// frameInterval: 每隔多少秒采样一帧,默认 30。
func GenerateVideoPreview(videoPath, outPath string, width, height int, frameInterval ...int) error {
v, err := NewVideo()
if err != nil {
return err
}
duration, err := getVideoDuration(videoPath)
if err != nil {
return err
}
ext := strings.ToLower(filepath.Ext(outPath))
vf := fmt.Sprintf("scale=%d:%d:force_original_aspect_ratio=increase,crop=%d:%d", width, height, width, height)
// 1. 单张图片模式
if ext == ".jpg" || ext == ".jpeg" || ext == ".png" {
t := duration * 0.5
cmd := exec.Command(v.FFmpegPath, "-ss", fmt.Sprintf("%f", t), "-i", videoPath, "-frames:v", "1", "-vf", vf, "-y", outPath)
return cmd.Run()
}
// 2. 动画或多图模式需要计算多帧
interval := 30
if len(frameInterval) > 0 && frameInterval[0] > 0 {
interval = frameInterval[0]
}
// 动态计算帧数: 避免过多的帧浪费 Token每 interval 秒 1 帧,最少 3 帧,最多 8 帧
frameCount := int(duration / float64(interval))
if frameCount < 3 {
frameCount = 3
} else if frameCount > 8 {
frameCount = 8
}
times := make([]float64, frameCount)
for i := 0; i < frameCount; i++ {
times[i] = duration * (0.10 + 0.80*(float64(i)/float64(frameCount-1)))
}
// 2a. 动画模式 (.webp, .gif)
if ext == ".webp" || ext == ".gif" {
tmpDir, _ := os.MkdirTemp("", "frames")
defer os.RemoveAll(tmpDir)
for i, t := range times {
framePath := filepath.Join(tmpDir, fmt.Sprintf("frame_%02d.png", i))
cmd := exec.Command(v.FFmpegPath, "-ss", fmt.Sprintf("%f", t), "-i", videoPath, "-frames:v", "1", "-vf", vf, "-y", framePath)
if err := cmd.Run(); err != nil {
return err
}
}
var cmd *exec.Cmd
if ext == ".webp" {
cmd = exec.Command(v.FFmpegPath, "-framerate", "1", "-i", filepath.Join(tmpDir, "frame_%02d.png"),
"-c:v", "libwebp", "-lossless", "0", "-quality", "70", "-loop", "0", "-y", outPath)
} else {
cmd = exec.Command(v.FFmpegPath, "-framerate", "1", "-i", filepath.Join(tmpDir, "frame_%02d.png"), "-y", outPath)
}
return cmd.Run()
}
// 2b. 文件夹多图模式
if err := os.MkdirAll(outPath, 0755); err != nil {
return err
}
for i, t := range times {
framePath := filepath.Join(outPath, fmt.Sprintf("%d.jpg", i+1))
cmd := exec.Command(v.FFmpegPath, "-ss", fmt.Sprintf("%f", t), "-i", videoPath, "-frames:v", "1", "-vf", vf, "-q:v", "2", "-y", framePath)
if err := cmd.Run(); err != nil {
return err
}
}
return nil
}
// GenerateAudioPreview 提取音频用于预览或语音转写
// 支持根据 outPath 后缀输出格式:
// - .ogg: 使用 libopus (16kHz, 单声道, 12kbps), 极致压缩且保留人声特征,适合转写
// - .wav: 标准 PCM (16kHz, 单声道), 无损但体积较大,部分转写引擎强制要求
// - 其他: 默认使用 libopus 转为 ogg
func GenerateAudioPreview(mediaPath, outPath string) error {
v, err := NewVideo()
if err != nil {
return err
}
ext := strings.ToLower(filepath.Ext(outPath))
// 通用参数: 禁用视频, 16kHz 采样率 (STT 标准), 单声道
args := []string{"-i", mediaPath, "-vn", "-ar", "16000", "-ac", "1"}
if ext == ".wav" {
// WAV 格式,保留 PCM最长 180 秒避免 LLM OOM
args = append(args, "-t", "180", "-y", outPath)
} else {
// 默认或 .ogg 使用 libopus 极致压缩,最长 180 秒
args = append(args, "-c:a", "libopus", "-b:a", "12k", "-t", "180", "-y", outPath)
}
cmd := exec.Command(v.FFmpegPath, args...)
return cmd.Run()
}
func getVideoDuration(videoPath string) (float64, error) {
out, err := exec.Command("ffprobe", "-v", "error", "-show_entries", "format=duration", "-of", "default=noprint_wrappers=1:nokey=1", videoPath).Output()
if err != nil {
return 0, err
}
var duration float64
_, err = fmt.Sscanf(strings.TrimSpace(string(out)), "%f", &duration)
return duration, err
}