audiogen, png2mp4

Signed-off-by: Balazs Horvath <acsipont@gmail.com>

audiogen_medium.py

@@ -0,0 +1,18 @@
+import sys
+import torchaudio
+from audiocraft.models import AudioGen
+from audiocraft.data.audio import audio_write
+
+model = AudioGen.get_pretrained('facebook/audiogen-medium')
+model.set_generation_params(duration=5)  # generate 5 seconds.
+descriptions = sys.argv[1:]
+if not descriptions:
+    print('At least one prompt should be provided')
+    sys.exit(1)
+wav = model.generate(descriptions)  # generates 3 samples.
+
+for one_wav, description in zip(wav, descriptions):
+    file_name = description.replace(' ', '_')
+    # Will save under {idx}.wav, with loudness normalization at -14 db LUFS.
+    audio_write(file_name, one_wav.cpu(), model.sample_rate, strategy="loudness", loudness_compressor=True)
+

zsh/png2mp4.zsh

@@ -10,81 +10,63 @@
 #
 # This function:
 # 1. Deactivates conda environment
-# 2. Creates a temporary directory for processing
-# 3. Finds all PNG files in the current directory
-# 4. Uses the current directory name as the output filename prefix
-# 5. Copies and optionally repeats images to the temp directory
-# 6. Uses ffmpeg to create an MP4 with x265 encoding, including:
+# 2. Finds all PNG files in the current directory
+# 3. Uses the current directory name as the output filename prefix
+# 4. Uses ffmpeg to create an MP4 with x265 encoding, including:
 #    - Frame rate of 60 fps
 #    - Image scaling to 1024x1024
 #    - Step count overlay text (divided by 1,000,000 and truncated to remove 6 zeros and decimal places)
 #    - High-quality encoding settings
-# 7. Adds padding and fade-out effect to the final video
-# 8. Cleans up temporary files
-# 9. Reactivates conda environment
+# 5. Adds padding and fade-out effect to the final video
+# 6. Reactivates conda environment
 #
 # Requirements:
 #   - ffmpeg with libx265 support
-#   - bc (basic calculator)
 #   - Inconsolata-Light.ttf font in /usr/share/fonts/TTF/
 png2mp4_x265() {
     conda deactivate
-    local max_images=""
     local step_multiplier=1
     local repeat=1
-    local temp_dir="/home/kade/.local/tmp"
+
     while [[ "$#" -gt 0 ]]; do
         case $1 in
-            --max) max_images="$2"; shift ;;
             --step) step_multiplier="$2"; shift ;;
             --repeat) repeat="$2"; shift ;;
             *) echo "Unknown parameter passed: $1"; return 1 ;;
         esac
         shift
     done
-    echo "Creating temporary directory..."
-    mkdir -p "$temp_dir"
-    echo "Checking for PNG files..."
-    png_files=($(/usr/bin/env ls *.png 2>/dev/null))
-    if [ ${#png_files[@]} -eq 0 ]; then
-        echo "Error: No PNG files found in the current directory."
-        return 1
-    fi
-    echo "Setting output filename..."
     output_filename="$(basename "$(pwd)")_x265.mp4"
     echo "Output filename: $output_filename"
-    echo "Creating repeated images..."
-    for img in "${png_files[@]}"; do
-        for i in $(seq 1 $repeat); do
-            cp "$img" "$temp_dir/${img%.*}_${i}.png"
-        done
-    done
+
+
+    local nframes=$(find . -type f -name '*.png' | wc -l)
+    local duration=$(($nframes * $repeat / 60.))
+    local fade_start=$((duration + 3))
+    echo "Found $nframes for a duration of $duration seconds"
+
+
     echo "Running ffmpeg with x265 encoding..."
-    if [[ -n "$max_images" ]]; then
-        dis_is_broken_dedupe_me
-        ffmpeg -framerate 60 -pattern_type glob -i "$temp_dir/*.png" -vf "scale=1024x1024,select='not(mod(n\,$max_images))',drawtext=fontfile=/usr/share/fonts/TTF/Inconsolata-Light.ttf:text='Steps\: %{eif\:trunc(n*$step_multiplier/$repeat)}':x=10:y=h-th-10:fontsize=24:fontcolor=white" \
-        -c:v libx265 -preset slower -x265-params "crf=18:qcomp=0.8:aq-mode=3:aq-strength=0.8:deblock=-1,-1" -pix_fmt yuv420p -y "$temp_dir/temp.mp4"
-    else
-        ffmpeg -framerate "60/$repeat" -pattern_type glob -i "*.png" -vf "scale=1024x1024,drawtext=fontfile=/usr/share/fonts/TTF/Inconsolata-Light.ttf:text='Steps\: %{eif\\:trunc(n*$step_multiplier/$repeat)\\:u\\:3}':x=10:y=h-th-10:fontsize=24:fontcolor=white" \
-        -c:v libx265 -preset slower -x265-params "crf=18:qcomp=0.8:aq-mode=3:aq-strength=0.8:deblock=-1,-1" -pix_fmt yuv420p -y "$temp_dir/temp.mp4"
-    fi
+    local font=/usr/share/fonts/TTF/Inconsolata-Light.ttf
+    local drawtext="drawtext=fontfile=${font}:text='Steps\: %{eif\\:trunc(n*$step_multiplier)\\:u\\:3}':x=10:y=h-th-10:fontsize=24:fontcolor=white"
+    local fadeout="tpad=stop_mode=clone:stop_duration=8,fade=t=out:st=${fade_start}:d=5"
+    local output_preset=(
+        -c:v libx265
+        -preset slower -tune animation
+        -g $repeat
+        -pix_fmt yuv420p
+        -x265-params "crf=20:qcomp=0.8:aq-mode=3:aq-strength=0.8:deblock=-1,-1"
+    )
+
+    ffmpeg -framerate "60/$repeat" -pattern_type glob -i "*.png" -r 60 -vf "scale=1024x1024,${drawtext},${fadeout}"\
+        -c:v ffv1 -f matroska - | ffmpeg -f matroska -i - \
+        -vf "$fadeout" "${output_preset[@]}" -y output.mp4
+
     if [ $? -ne 0 ]; then
         echo "Error: ffmpeg command failed."
-        rm -rf "$temp_dir"
-        return 1
-    fi
-    echo "Processing final video..."
-    duration=$(ffmpeg -i "$temp_dir/temp.mp4" 2>&1 | grep 'Duration' | awk '{print $2}' | tr -d , | awk -F: '{print ($1 * 3600) + ($2 * 60) + $3}')
-    fade_start=$(echo "$duration + 3" | bc)
-    ffmpeg -i "$temp_dir/temp.mp4" -vf "tpad=stop_mode=clone:stop_duration=8,fade=t=out:st=$fade_start:d=5" \
-    -c:v libx265 -preset slower -x265-params "crf=18:qcomp=0.8:aq-mode=3:aq-strength=0.8:deblock=-1,-1" -pix_fmt yuv420p -y "$output_filename"
-    if [ $? -ne 0 ]; then
-        echo "Error: Final ffmpeg processing failed."
-        rm -rf "$temp_dir"
         return 1
     fi
-    echo "Cleaning up temporary files..."
-    rm -rf "$temp_dir"
+
     conda activate
     echo "Process completed successfully."
 }