fix: use ffprobe for SAR detection and calculate zscale width

internal/encoder/encoder.go

@@ -1,6 +1,7 @@
 package encoder
 
 import (
+	"encoding/json"
 	"fmt"
 	"os"
 	"os/exec"
@@ -13,12 +14,26 @@ import (
 
 type Encoder struct {
 	ffmpegPath  string
+	ffprobePath string
 	opusencPath string
 }
 
+type StreamInfo struct {
+	Width              int    `json:"width"`
+	Height             int    `json:"height"`
+	SampleAspectRatio  string `json:"sample_aspect_ratio"`
+	DisplayAspectRatio string `json:"display_aspect_ratio"`
+	CodecName          string `json:"codec_name"`
+}
+
+type ProbeResult struct {
+	Streams []StreamInfo `json:"streams"`
+}
+
 func New() *Encoder {
 	return &Encoder{
 		ffmpegPath:  "ffmpeg",
+		ffprobePath: "ffprobe",
 		opusencPath: "opusenc",
 	}
 }
@@ -27,6 +42,9 @@ func (e *Encoder) CheckDeps() error {
 	if _, err := exec.LookPath(e.ffmpegPath); err != nil {
 		return fmt.Errorf("ffmpeg not found")
 	}
+	if _, err := exec.LookPath(e.ffprobePath); err != nil {
+		return fmt.Errorf("ffprobe not found")
+	}
 	if _, err := exec.LookPath(e.opusencPath); err != nil {
 		return fmt.Errorf("opusenc not found")
 	}
@@ -34,29 +52,83 @@ func (e *Encoder) CheckDeps() error {
 }
 
 func (e *Encoder) GetMediaInfo(path string) (width, height int, interlaced bool, err error) {
-	cmd := exec.Command(e.ffmpegPath, "-i", path, "-hide_banner", "-vf", "idet")
+	cmd := exec.Command(e.ffprobePath, "-v", "error", "-show_streams", "-print_format", "json", path)
-	output, _ := cmd.StderrPipe()
+	output, err := cmd.CombinedOutput()
+	if err != nil {
+		return 0, 0, false, fmt.Errorf("ffprobe error: %w", err)
+	}
+
+	fmt.Printf("DEBUG ffprobe output: %s\n", string(output))
+
+	var result ProbeResult
+	if err := json.Unmarshal(output, &result); err != nil {
+		return 0, 0, false, fmt.Errorf("parsing ffprobe output: %w", err)
+	}
+
+	for _, stream := range result.Streams {
+		if stream.CodecName == "mpeg2video" || stream.CodecName == "h264" || stream.CodecName == "hevc" {
+			width = stream.Width
+			height = stream.Height
+			fmt.Printf("DEBUG detected video: %dx%d, SAR=%s\n", width, height, stream.SampleAspectRatio)
+			break
+		}
+	}
+
+	cmd = exec.Command(e.ffmpegPath, "-i", path, "-hide_banner", "-vf", "idet")
+	stderr, _ := cmd.StderrPipe()
 	cmd.Start()
-
 	buf := make([]byte, 16384)
-	n, _ := output.Read(buf)
+	n, _ := stderr.Read(buf)
-	output.Close()
+	stderr.Close()
-
 	outputStr := string(buf[:n])
+	interlaced = strings.Contains(outputStr, "TFF") || strings.Contains(outputStr, "BFF")
 
-	var w, h int
+	fmt.Printf("DEBUG interlaced: %v\n", interlaced)
-	for i := 0; i < len(outputStr)-10; i++ {
+
-		if outputStr[i] >= '0' && outputStr[i] <= '9' {
+	return width, height, interlaced, nil
-			fmt.Sscanf(outputStr[i:], "%dx%d", &w, &h)
+}
-			if w > 0 && h > 0 {
+
-				break
+func (e *Encoder) calculateZscaleWidth(path string, originalHeight int) (string, int, error) {
+	cmd := exec.Command(e.ffprobePath, "-v", "error", "-select_streams", "v:0", "-show_entries", "stream=width,height,sample_aspect_ratio", "-print_format", "json", path)
+	output, err := cmd.CombinedOutput()
+	if err != nil {
+		return "", 0, fmt.Errorf("ffprobe error: %w", err)
+	}
+
+	fmt.Printf("DEBUG ffprobe for zscale: %s\n", string(output))
+
+	var result ProbeResult
+	if err := json.Unmarshal(output, &result); err != nil {
+		return "", 0, fmt.Errorf("parsing ffprobe: %w", err)
+	}
+
+	if len(result.Streams) == 0 {
+		return "", 0, fmt.Errorf("no video streams found")
+	}
+
+	stream := result.Streams[0]
+	width := stream.Width
+	height := stream.Height
+
+	sar := stream.SampleAspectRatio
+	fmt.Printf("DEBUG stream: width=%d, height=%d, sar=%s\n", width, height, sar)
+
+	newWidth := width
+
+	if sar != "1:1" && sar != "N/A" && sar != "" {
+		parts := strings.Split(sar, ":")
+		if len(parts) == 2 {
+			num, err1 := strconv.Atoi(parts[0])
+			den, err2 := strconv.Atoi(parts[1])
+			if err1 == nil && err2 == nil && den != 0 {
+				newWidth = (width * num) / den
+				fmt.Printf("DEBUG calculated new width: %d (sar=%s)\n", newWidth, sar)
 			}
 		}
 	}
 
-	interlaced = strings.Contains(outputStr, "TFF") || strings.Contains(outputStr, "BFF")
+	zscale := fmt.Sprintf("zscale=w=%d:h=%d:filter=spline36", newWidth, height)
-
+	return zscale, newWidth, nil
-	return w, h, interlaced, nil
 }
 
 func (e *Encoder) Transcode(input string, job *types.Job, metadata *types.Metadata, interlaced bool) error {
@@ -111,26 +183,35 @@ func (e *Encoder) encodeVideo(input string, opusFiles []string, job *types.Job,
 
 	svtParams := fmt.Sprintf("film-grain=10:film-grain-denoise=1:scd=1:qm-min=4:qm-max=15:preset=%d", job.Preset)
 
-	vf := "zscale=filter=spline36:mode=16:9"
+	zscaleStr, newWidth, err := e.calculateZscaleWidth(input, 0)
+	if err != nil {
+		return fmt.Errorf("calculating zscale: %w", err)
+	}
+
+	fmt.Printf("DEBUG zscale: %s (newWidth=%d)\n", zscaleStr, newWidth)
+
+	vf := zscaleStr
 	if interlaced {
 		vf = "bwdif=mode=0:par=-1:-1," + vf
 	}
 
+	fmt.Printf("DEBUG final vf: %s\n", vf)
+
 	args := []string{
 		"-y",
+		"-i", input,
 	}
 
-	args = append(args, "-i", input)
-	args = append(args, "-c:v", "libsvtav1")
-	args = append(args, "-crf", strconv.Itoa(job.CRF))
-	args = append(args, "-svtav1-params", svtParams)
-	args = append(args, "-pix_fmt", "yuv420p10le")
-	args = append(args, "-filter:v:0", vf)
-
 	for _, opus := range opusFiles {
 		args = append(args, "-i", opus)
 	}
 
+	args = append(args, "-vf", vf)
+	args = append(args, "-c:v", "libsvtav1")
+	args = append(args, "-crf", strconv.Itoa(job.CRF))
+	args = append(args, "-svtav1-params", svtParams)
+	args = append(args, "-pix_fmt", "yuv420p10le")
+
 	args = append(args, "-map", "0:v")
 	args = append(args, "-map", "0:s?")
 
@@ -158,7 +239,9 @@ func (e *Encoder) encodeVideo(input string, opusFiles []string, job *types.Job,
 
 	args = append(args, outFile)
 
+	args = append([]string{"-hide_banner", "-v", "error"}, args...)
 	cmd := exec.Command(e.ffmpegPath, args...)
+	fmt.Printf("DEBUG FFmpeg command: ffmpeg %s\n", strings.Join(args, " "))
 	if out, err := cmd.CombinedOutput(); err != nil {
 		return fmt.Errorf("ffmpeg encode: %s %w", out, err)
 	}