qual-bench.sh : add quality comparison tool, and update main.cpp to allow using a font file (#569)

examples/main/main.cpp

@@ -80,6 +80,7 @@ struct whisper_params {
 
     std::string language = "en";
     std::string prompt;
+    std::string font_path = "/System/Library/Fonts/Supplemental/Courier New Bold.ttf";
     std::string model    = "models/ggml-base.en.bin";
 
     std::vector<std::string> fname_inp = {};
@@ -127,6 +128,7 @@ bool whisper_params_parse(int argc, char ** argv, whisper_params & params) {
         else if (arg == "-ovtt" || arg == "--output-vtt")     { params.output_vtt     = true; }
         else if (arg == "-osrt" || arg == "--output-srt")     { params.output_srt     = true; }
         else if (arg == "-owts" || arg == "--output-words")   { params.output_wts     = true; }
+        else if (arg == "-fp"   || arg == "--font-path")      { params.font_path      = argv[++i]; }
         else if (arg == "-ocsv" || arg == "--output-csv")     { params.output_csv     = true; }
         else if (arg == "-of"   || arg == "--output-file")    { params.fname_out.emplace_back(argv[++i]); }
         else if (arg == "-ps"   || arg == "--print-special")  { params.print_special  = true; }
@@ -174,6 +176,7 @@ void whisper_print_usage(int /*argc*/, char ** argv, const whisper_params & para
     fprintf(stderr, "  -ovtt,     --output-vtt        [%-7s] output result in a vtt file\n",                    params.output_vtt ? "true" : "false");
     fprintf(stderr, "  -osrt,     --output-srt        [%-7s] output result in a srt file\n",                    params.output_srt ? "true" : "false");
     fprintf(stderr, "  -owts,     --output-words      [%-7s] output script for generating karaoke video\n",     params.output_wts ? "true" : "false");
+    fprintf(stderr, "  -fp,       --font-path         [%-7s] path to a monospace font for karaoke video\n",     params.font_path.c_str());
     fprintf(stderr, "  -ocsv,     --output-csv        [%-7s] output result in a CSV file\n",                    params.output_csv ? "true" : "false");
     fprintf(stderr, "  -of FNAME, --output-file FNAME [%-7s] output file path (without file extension)\n",      "");
     fprintf(stderr, "  -ps,       --print-special     [%-7s] print special tokens\n",                           params.print_special ? "true" : "false");
@@ -368,13 +371,18 @@ bool output_csv(struct whisper_context * ctx, const char * fname) {
 // karaoke video generation
 // outputs a bash script that uses ffmpeg to generate a video with the subtitles
 // TODO: font parameter adjustments
-bool output_wts(struct whisper_context * ctx, const char * fname, const char * fname_inp, const whisper_params & /*params*/, float t_sec) {
+bool output_wts(struct whisper_context * ctx, const char * fname, const char * fname_inp, const whisper_params & params, float t_sec) {
     std::ofstream fout(fname);
 
     fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);
 
-    // TODO: become parameter
+    static const char * font = params.font_path.c_str();
-    static const char * font = "/System/Library/Fonts/Supplemental/Courier New Bold.ttf";
+
+    std::ifstream fin(font);
+    if (!fin.is_open()) {
+        fprintf(stderr, "%s: font not found at '%s', please specify a monospace font with -fp\n", __func__, font);
+        return false;
+    }
 
     fout << "#!/bin/bash" << "\n";
     fout << "\n";

extra/qual-bench.sh

@@ -0,0 +1,67 @@
+# This script takes two arguments
+# - an audio file
+# - [optional] path to a font file 
+
+# I'm using "/usr/share/fonts/truetype/freefont/FreeMono.ttf" on Ubuntu
+
+if [ -z "$1" ]; then
+    echo "Usage: $0 <audio file> [font file]"
+    exit 1
+fi
+
+#TODO: Make this a command line parameter
+models="base small large"
+#models="tiny.en tiny base.en base small.en small medium.en medium large-v1 large"
+
+DURATION=$(ffprobe -i $1 -show_entries format=duration -v quiet -of csv="p=0")
+DURATION=$(printf "%.2f" $DURATION)
+echo "Input file duration: ${DURATION}s"
+
+for model in $models; do
+    echo "Running $model"
+    COMMAND="./main -m models/ggml-$model.bin -owts -f $1 -of $1.$model" 
+
+    if [ ! -z "$2" ]; then
+        COMMAND="$COMMAND -fp $2"
+    fi
+    #TODO: Surface errors better
+    # TIMEFMT is for zsh, TIMEFORMAT is for bash
+    EXECTIME=$({ TIMEFMT="%E";TIMEFORMAT=%E; time $COMMAND >/dev/null 2>&1; } 2>&1)
+
+    # Slightly different formats between zsh and bash
+    if [ "${EXECTIME: -1}" == "s" ]; then
+        EXECTIME=${EXECTIME::-1}
+    fi
+
+    RATIO=$(echo "$DURATION / $EXECTIME" | bc -l)
+    RATIO=$(printf "%.2f" $RATIO)
+
+    echo "Execution time: ${EXECTIME}s (${RATIO}x realtime)"
+
+    # If the file already exists, delete it
+    if [ -f $1.mp4 ]; then
+        rm $1.$model.mp4
+    fi
+
+    bash $1.$model.wts >/dev/null 2>&1
+    mv $1.mp4 $1.$model.mp4
+
+    ffmpeg -y -f lavfi -i color=c=black:s=1200x50:d=$DURATION -vf "drawtext=fontfile=$2:fontsize=36:x=10:y=(h-text_h)/2:text='ggml-$model - ${EXECTIME}s (${RATIO}x realtime)':fontcolor=lightgrey" $1.$model.info.mp4 >/dev/null 2>&1
+done
+
+COMMAND="ffmpeg -y"
+for model in $models; do
+    COMMAND="$COMMAND -i $1.$model.info.mp4 -i $1.$model.mp4"
+done
+COMMAND="$COMMAND -filter_complex \""
+COUNT=0
+for model in $models; do
+    COMMAND="$COMMAND[${COUNT}:v][$(($COUNT+1)):v]"
+    COUNT=$((COUNT+2))
+done
+COMMAND="$COMMAND vstack=inputs=${COUNT}[v]\" -map \"[v]\" -map 1:a $1.all.mp4 >/dev/null 2>&1"
+
+echo $COMMAND
+
+# Run the command
+eval $COMMAND