vision/preview.go

120 lines
3.7 KiB
Go
Raw Permalink 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)
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)
if strings.HasSuffix(strings.ToLower(outPath), ".webp") {
// 借用 FFmpeg 将生成的画布转为高质量 WebP
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 (动态 WebP)
// 支持缩放并裁剪以填充指定尺寸 (Fill 模式)
func GenerateVideoPreview(videoPath, outPath string, width, height int) error {
v, err := NewVideo()
if err != nil {
return err
}
duration, err := getVideoDuration(videoPath)
if err != nil {
return err
}
// 动态计算帧数: 适合交给大模型 (VLM) 处理
// 避免过多的帧浪费 Token每 30 秒 1 帧,最少 3 帧,最多 8 帧
frameCount := int(duration / 30.0)
if frameCount < 3 {
frameCount = 3
} else if frameCount > 8 {
frameCount = 8
}
// 在 10% 到 90% 之间均匀采样,跳过片头片尾的可能黑屏
times := make([]float64, frameCount)
if frameCount == 1 {
times[0] = duration * 0.5
} else {
for i := 0; i < frameCount; i++ {
times[i] = duration * (0.10 + 0.80*(float64(i)/float64(frameCount-1)))
}
}
tmpDir, _ := os.MkdirTemp("", "frames")
defer os.RemoveAll(tmpDir)
// 使用 ffmpeg 的 scale 和 crop 滤镜实现 Fill 效果
vf := fmt.Sprintf("scale=%d:%d:force_original_aspect_ratio=increase,crop=%d:%d", width, height, width, height)
for i, t := range times {
framePath := filepath.Join(tmpDir, fmt.Sprintf("frame_%02d.png", i))
// 使用 -ss 快速 seek性能极高
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
}
}
// 生成 WebP 动画,每秒 1 帧 (1000ms 间隔,让大模型能看清每一帧)
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)
return cmd.Run()
}
// GenerateAudioPreview 提取 3 分钟内的音频用于预览或语音转写
// 格式: Ogg Opus, 16kHz, 单声道, 12kbps (极致压缩,保留人声特征)
func GenerateAudioPreview(mediaPath, outPath string) error {
v, err := NewVideo()
if err != nil {
return err
}
// -vn: 禁用视频
// -c:a libopus: 高效音频压缩
// -ar 16000: 采样率 16k (转写标准)
// -ac 1: 单声道
// -b:a 12k: 极致压缩
// -t 180: 最长 180 秒 (足以获得内容概要)
cmd := exec.Command(v.FFmpegPath, "-i", mediaPath, "-vn", "-c:a", "libopus", "-ar", "16000", "-ac", "1", "-b:a", "12k", "-t", "180", "-y", outPath)
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
}