easyai-ai-gateway/apps/api/internal/runner/video_duration.go

145 lines
3.7 KiB
Go

package runner
import (
"context"
"encoding/json"
"fmt"
"math"
"os/exec"
"strconv"
"strings"
"time"
)
const generatedVideoMetadataProbeTimeout = 8 * time.Second
type generatedVideoMetadata struct {
Duration float64
HasAudio bool
HasAudioKnown bool
}
type ffprobeVideoMetadata struct {
Format struct {
Duration string `json:"duration"`
} `json:"format"`
Streams []struct {
CodecType string `json:"codec_type"`
} `json:"streams"`
}
func (s *Service) enrichGeneratedVideoMetadata(ctx context.Context, taskKind string, result map[string]any) map[string]any {
if taskKind != "videos.generations" {
return result
}
data, _ := result["data"].([]any)
if len(data) == 0 {
return result
}
for _, raw := range data {
item, _ := raw.(map[string]any)
if len(item) == 0 || !isGeneratedVideoItem(item) {
continue
}
needsDuration := floatFromAny(item["duration"]) <= 0
_, hasAudioMetadata := boolishOptional(firstPresentValue(item, "has_audio", "hasAudio"))
if !needsDuration && hasAudioMetadata {
continue
}
urlValue := firstNonEmptyStringValue(item, "video_url", "videoUrl", "url")
if urlValue == "" {
continue
}
metadata, err := s.probeVideoMetadata(ctx, urlValue)
if err != nil {
if s.logger != nil {
s.logger.Debug("probe generated video metadata failed", "url", trimForLog(urlValue), "error", err)
}
continue
}
if needsDuration && metadata.Duration > 0 {
item["duration"] = metadata.Duration
}
if !hasAudioMetadata && metadata.HasAudioKnown {
item["has_audio"] = metadata.HasAudio
}
}
return result
}
func isGeneratedVideoItem(item map[string]any) bool {
itemType := strings.TrimSpace(stringFromAny(item["type"]))
if itemType == "video" {
return true
}
if firstNonEmptyStringValue(item, "video_url", "videoUrl") != "" {
return true
}
urlValue := strings.ToLower(firstNonEmptyStringValue(item, "url"))
return strings.Contains(urlValue, ".mp4") ||
strings.Contains(urlValue, ".mov") ||
strings.Contains(urlValue, ".webm") ||
strings.Contains(urlValue, ".m3u8")
}
func (s *Service) probeVideoMetadata(ctx context.Context, rawURL string) (generatedVideoMetadata, error) {
if _, err := exec.LookPath("ffprobe"); err != nil {
return generatedVideoMetadata{}, err
}
probeURL := rawURL
if s != nil {
if resolved, err := s.generatedAssetFetchURL(rawURL); err == nil && strings.TrimSpace(resolved) != "" {
probeURL = resolved
}
}
probeCtx, cancel := context.WithTimeout(ctx, generatedVideoMetadataProbeTimeout)
defer cancel()
cmd := exec.CommandContext(
probeCtx,
"ffprobe",
"-v", "error",
"-show_entries", "format=duration:stream=codec_type",
"-of", "json",
probeURL,
)
output, err := cmd.Output()
if err != nil {
return generatedVideoMetadata{}, err
}
var probed ffprobeVideoMetadata
if err := json.Unmarshal(output, &probed); err != nil {
return generatedVideoMetadata{}, err
}
metadata := generatedVideoMetadata{}
if durationText := strings.TrimSpace(probed.Format.Duration); durationText != "" {
if duration, err := strconv.ParseFloat(durationText, 64); err == nil && duration > 0 && !math.IsNaN(duration) && !math.IsInf(duration, 0) {
rounded := math.Round(duration)
if rounded <= 0 {
rounded = 1
}
metadata.Duration = rounded
}
}
if probed.Streams != nil {
metadata.HasAudioKnown = true
for _, stream := range probed.Streams {
if strings.TrimSpace(stream.CodecType) == "audio" {
metadata.HasAudio = true
break
}
}
}
if metadata.Duration <= 0 && !metadata.HasAudioKnown {
return metadata, fmt.Errorf("invalid video metadata: %q", trimForLog(string(output)))
}
return metadata, nil
}
func trimForLog(value string) string {
value = strings.TrimSpace(value)
if len(value) <= 120 {
return value
}
return value[:120] + "..."
}