refactor: abstraction for transcoding

go/internal/media/helpers.go

@@ -0,0 +1,107 @@
+package media
+
+import (
+	"bytes"
+	"context"
+	"errors"
+	"log/slog"
+	"os"
+	"os/exec"
+	"strings"
+)
+
+func IsIOSPlayableMime(mime string) bool {
+	switch mime {
+	case "video/mp4", "video/quicktime", "audio/mp4", "audio/aac", "audio/x-m4a", "audio/mpeg":
+		return true
+	}
+	return false
+}
+
+func IsAudio(mime string) bool {
+	return strings.HasPrefix(mime, "audio/")
+}
+
+type TranscodeInput struct {
+	SourceURL string
+	MimeType  string
+}
+
+type TranscodeOutput struct {
+	TempLocalFilePath string
+	OutputMimeType    string
+	// Extension such as ".m4a" or ".mp4"
+	OutputExt string
+}
+
+var (
+	ErrInvalidInput error = errors.New("invalid input")
+)
+
+// TranscodeToMp4 takes in any audio or video source URL and
+// returns the filepath of the transcoded media
+// WARNING: caller responsible for deleting TempLocalFilePath
+func TranscodeToMp4(ctx context.Context, input TranscodeInput) (*TranscodeOutput, error) {
+	if input.SourceURL == "" || input.MimeType == "" {
+		return nil, ErrInvalidInput
+	}
+
+	isAudio := IsAudio(input.MimeType)
+	var outputExt, outputMime string
+	if isAudio {
+		outputExt = ".m4a"
+		outputMime = "audio/mp4"
+	} else {
+		outputExt = ".mp4"
+		outputMime = "video/mp4"
+	}
+
+	tmp, err := os.CreateTemp("", "transcode-*"+outputExt)
+	if err != nil {
+		slog.Error("transcode: failed to create temp file", "error", err)
+		return nil, err
+	}
+	tmpPath := tmp.Name()
+	tmp.Close()
+
+	var args []string
+	if isAudio {
+		args = []string{
+			"-y", "-i", input.SourceURL,
+			"-vn",
+			"-c:a", "aac", "-b:a", "128k",
+			"-movflags", "+faststart",
+			tmpPath,
+		}
+	} else {
+		// Cap encoder parallelism and lookahead to keep memory bounded — screen
+		// recordings come in at native display resolution (often 1440p–4K) and
+		// libx264's per-thread lookahead/reference buffers blow past the worker's
+		// memory limit otherwise. Output is also downscaled to 1080p max, which
+		// mobile playback won't notice; the original WebM stays in GCS untouched.
+		args = []string{
+			"-y", "-i", input.SourceURL,
+			"-vf", "scale='min(1920,iw)':-2:flags=lanczos",
+			"-c:v", "libx264", "-preset", "veryfast", "-crf", "23",
+			"-pix_fmt", "yuv420p", "-profile:v", "baseline", "-level", "3.1",
+			"-x264-params", "rc-lookahead=20:ref=2",
+			"-threads", "2", "-filter_threads", "2",
+			"-c:a", "aac", "-b:a", "128k",
+			"-movflags", "+faststart",
+			tmpPath,
+		}
+	}
+
+	cmd := exec.CommandContext(ctx, "ffmpeg", args...)
+	var stderr bytes.Buffer
+	cmd.Stderr = &stderr
+	if err := cmd.Run(); err != nil {
+		return nil, errors.Join(err, errors.New(stderr.String()))
+	}
+
+	return &TranscodeOutput{
+		TempLocalFilePath: tmpPath,
+		OutputMimeType:    outputMime,
+		OutputExt:         outputExt,
+	}, nil
+}