IDK man, changes I guess

export.go

@@ -41,14 +41,20 @@ type ExportRequest struct {
 	// animated GIF (also uses Height and FPS)
 	Dither string `json:"dither"` // one of gifDithers
 	Loop   bool   `json:"loop"`
+	// captions burned into the picture (re-encode and GIF only)
+	Captions     map[string]string `json:"captions"`     // source id -> .srt path
+	CaptionSize  string            `json:"captionSize"`  // key of captionSizes
+	CaptionStyle string            `json:"captionStyle"` // key of captionStyles
 }
 
 // gifDithers are the paletteuse dither modes offered for GIF export.
 var gifDithers = []string{"sierra2_4a", "bayer", "none"}
 
 type clipJob struct {
-	m       *Media
-	in, out float64
+	m        *Media
+	in, out  float64
+	cues     []Cue  // captions to burn in, on the source's timeline
+	captions string // .srt file with the clip's cues, timed from In (see writeClipCaptions)
 }
 
 func (c clipJob) dur() float64 { return c.out - c.in }
@@ -109,6 +115,10 @@ func (a *App) handleExport(w http.ResponseWriter, r *http.Request) {
 		writeError(w, http.StatusBadRequest, "unknown export mode")
 		return
 	}
+	if err := a.attachCaptions(&req, clips); err != nil {
+		writeError(w, http.StatusBadRequest, err.Error())
+		return
+	}
 
 	outputs := []string{out}
 	if !req.Join {
@@ -334,6 +344,14 @@ func (a *App) runExport(ctx context.Context, c *JobCtl, req *ExportRequest, clip
 		}
 	}
 
+	captionDir, err := writeClipCaptions(clips)
+	if captionDir != "" {
+		defer os.RemoveAll(captionDir)
+	}
+	if err != nil {
+		return fmt.Errorf("could not prepare the captions: %w", err)
+	}
+
 	switch {
 	case copyJoin:
 		out := outputs[0]
@@ -527,6 +545,7 @@ func encodeArgs(clips []clipJob, req *ExportRequest, ext string) ([]string, erro
 	}
 	anyAudio = anyAudio && !req.NoAudio
 
+	f.CaptionSize, f.CaptionStyle = req.CaptionSize, req.CaptionStyle
 	vc, _ := codecByName(req.VCodec)
 	var venc []string
 	venc, f.PixFmt = videoEncoderArgs(vc, req.Quality, req.Speed)
@@ -567,6 +586,7 @@ func gifArgs(clips []clipJob, req *ExportRequest, palette string) (paletteRun, g
 		return nil, nil, err
 	}
 	f.PixFmt = "bgra" // what palettegen and paletteuse take
+	f.CaptionSize, f.CaptionStyle = req.CaptionSize, req.CaptionStyle
 	inputs, graph, n := joinGraph(clips, f)
 
 	paletteRun = append(ffmpegBaseArgs(), inputs...)
@@ -593,6 +613,8 @@ type outFormat struct {
 	PixFmt     string
 	SampleRate int
 	Layout     string
+	// how burned-in captions look (keys of captionSizes and captionStyles)
+	CaptionSize, CaptionStyle string
 }
 
 // frameFormat picks the output frame size and rate: the first video clip's
@@ -625,7 +647,8 @@ func frameFormat(clips []clipJob, height int, fps string) (outFormat, error) {
 // joinGraph returns the input arguments (one seeked input per clip, plus
 // generated black video or silence for clips that lack it), how many inputs
 // that is, and a filter graph that converts every clip to f, letterboxing as
-// needed, and joins them with the concat filter into [outv] (and [outa]).
+// needed and drawing its captions, and joins them with the concat filter into
+// [outv] (and [outa]).
 func joinGraph(clips []clipJob, f outFormat) (args []string, graph string, inputs int) {
 	var filters []string
 	var concatIn strings.Builder
@@ -637,17 +660,24 @@ func joinGraph(clips []clipJob, f outFormat) (args []string, graph string, input
 		src := inputs
 		inputs++
 
+		// Captions are drawn on the output-sized frame, so they get the same
+		// size in every clip. After fps the frames are timed from the In point
+		// on the output frame grid, which is how writeClipCaptions times them.
+		captions := ""
+		if cl.captions != "" {
+			captions = "," + captionFilter(cl.captions, f)
+		}
 		if v := cl.m.Video; v != nil {
 			chain := fmt.Sprintf("[%d:%d]", src, v.Index)
 			if !v.SquarePixels {
 				chain += "scale='trunc(iw*sar/2)*2':ih,setsar=1,"
 			}
-			chain += fmt.Sprintf("scale=%d:%d:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=%d:%d:(ow-iw)/2:(oh-ih)/2:color=black,setsar=1,fps=%s,format=%s[v%d]",
-				f.W, f.H, f.W, f.H, f.FPS, f.PixFmt, i)
+			chain += fmt.Sprintf("scale=%d:%d:force_original_aspect_ratio=decrease:force_divisible_by=2,pad=%d:%d:(ow-iw)/2:(oh-ih)/2:color=black,setsar=1,fps=%s%s,format=%s[v%d]",
+				f.W, f.H, f.W, f.H, f.FPS, captions, f.PixFmt, i)
 			filters = append(filters, chain)
 		} else {
 			args = append(args, "-f", "lavfi", "-t", d, "-i", fmt.Sprintf("color=c=black:s=%dx%d:r=%s", f.W, f.H, f.FPS))
-			filters = append(filters, fmt.Sprintf("[%d:v]setsar=1,format=%s[v%d]", inputs, f.PixFmt, i))
+			filters = append(filters, fmt.Sprintf("[%d:v]setsar=1%s,format=%s[v%d]", inputs, captions, f.PixFmt, i))
 			inputs++
 		}
 		fmt.Fprintf(&concatIn, "[v%d]", i)