You'll get even smaller files. I've been replacing my media library with AV1 when I see titles available. Still nice though. I used to have a script to automatically download all videos from a few youtube channels. Would have been useful to combine it with this.

#!/usr/bin/env bash # ======================================== # INTELLIGENT HEVC VIDEO COMPRESSION PIPELINE # ---------------------------------------------------------------------------- # This script compresses video files to H.265 (HEVC) with Opus audio. # It is designed for archival purposes where file size reduction is # prioritized while maintaining perceptual quality. # # Features: # - Two-pass or Single-pass libx265 encoding with smart bitrate targeting. # - Automatic audio normalization (Peak detection -> Target dB). # - Black bar detection and cropping. # - Logarithmic bitrate targeting (soft compression for high bitrate sources) # OR Manual scaling factor targeting (strict size reduction). # # Usage: # ./script.sh [scale_factor] <input_file> # # Examples: # ./script.sh video.mkv (Use auto bitrate curve) # ./script.sh 2.0 video.mkv (Target 1/2 of original size) # ./script.sh 1.5 video.mkv (Target 1/1.5 of original size) # # Author: Clort (clort81) # Requires: ffmpeg, ffprobe, bc, coreutils # ======================================== # Strict Mode: # -e: Exit immediately if a command exits with a non-zero status. # -u: Treat unset variables as an error. # -o pipefail: Return value of a pipeline is the status of the last command to exit with a non-zero status. set -euo pipefail # ---------------------------- # GLOBAL CONSTANTS # ---------------------------- readonly MAX_WIDTH=960 # Target Width (Quarter HD) readonly MAX_HEIGHT=540 # Target Height readonly VIDEO_CODEC="libx265" # HEVC Encoder readonly AUDIO_BITRATE="12k" # Opus Bitrate (Speech optimized) readonly ENCODING_PRESET="fast" # Speed/Efficiency tradeoff # Audio Normalization Target (dBFS) # The script detects the current maximum volume and calculates gain # to reach this target level. # Example: -20 targets -20dB. 0 targets 0dB (Maximum digital volume). readonly TARGET_VOL="-18" # Encoding Parameters for Single Pass # ref=4: Decent reference frames for compression without killing speed. # subme=6: High quality motion estimation. # aq-mode=1: Adaptive Quantization (varies bitrate based on frame complexity). readonly ENCODING_PARAMS="scc=2:weightp=1:weightb=1:ref=4:bframes=8:rc-lookahead=60:aq-mode=1:deblock=-3,-3:subme=6:rd=4" # Scaling Filter # force_original_aspect_ratio=decrease ensures we fit within the box without stretching. readonly SCALE_FILTER="scale='min(${MAX_WIDTH},iw):min(${MAX_HEIGHT},ih)':force_original_aspect_ratio=decrease" # Bitrate Mapping Constants (Auto-Mode only) readonly MIN_TARGET_KBPS=110 readonly MAX_TARGET_KBPS=192 # ---------------------------- # HELPER FUNCTIONS # ---------------------------- # Usage: die "Error message" die() { echo "ERROR: $*" >&2 exit 1 } # Get duration in seconds (float) # Relies on ffprobe returning standard format=duration get_duration() { local file="$1" local dur dur=$(ffprobe -v error -show_entries format=duration -of csv=p=0 "$file" 2>/dev/null) || return 1 case "$dur" in ""|"N/A") return 1 ;; esac printf '%s\n' "$dur" } # Get video stream bitrate (bps) get_stream_bitrate_bps() { local file="$1" local br # -select_streams v:0 selects only the first video stream br=$(ffprobe -v quiet -select_streams v:0 -show_entries stream=bit_rate -of csv=p=0 "$file" 2>/dev/null) || return 1 [[ "$br" =~ ^[0-9]+$ ]] && (( br > 0 )) && echo "$br" } # Get global container bitrate (bps) get_global_bitrate_bps() { local file="$1" local br br=$(ffprobe -v quiet -show_entries format=bit_rate -of csv=p=0 "$file" 2>/dev/null) || return 1 [[ "$br" =~ ^[0-9]+$ ]] && (( br > 0 )) && echo "$br" } # Fallback: compute bitrate from file size / duration compute_bitrate_from_filesize() { local file="$1" local dur dur=$(get_duration "$file") || return 1 if ! printf '%s\n' "$dur" | grep -E '^[0-9]*\.?[0-9]+$' >/dev/null; then return 1 fi if (( $(bc -l <<< "$dur <= 0") )); then return 1 fi local size # stat -c%s returns size in bytes (Linux standard) size=$(stat -c%s "$file" 2>/dev/null) || return 1 (( size > 0 )) || return 1 local br_bps br_bps=$(bc -l <<< "($size * 8) / $dur" 2>/dev/null) || return 1 br_bps=$(printf "%.0f" "$br_bps" 2>/dev/null) [[ -n "$br_bps" ]] && (( br_bps > 0 )) && echo "$br_bps" } # Determine best estimate for input bitrate get_effective_video_bitrate_bps() { local file="$1" echo " -> Attempting stream-level bitrate extraction..." >&2 if get_stream_bitrate_bps "$file"; then return 0; fi echo " -> Falling back to global container bitrate..." >&2 if get_global_bitrate_bps "$file"; then return 0; fi echo " -> Falling back to file size / duration calculation..." >&2 if compute_bitrate_from_filesize "$file"; then return 0; fi die "All methods failed to determine input bitrate for $file" } # Log-curve calculation for Auto-Mode # Formula: y = 488.4 * ln(x + 1000) - 3211 calculate_target_bitrate_kbps() { local x="$1" # Sanitize: ensure we only pass numbers to bc x=$(printf '%s' "$x" | tr -cd '0-9.') [[ -z "$x" || "$x" == "." ]] && x=800 local y # bc logic: calculate y, then scale=0 to round to nearest integer y=$(bc -l <<< " scale = 10 x = $x if (x + 1000 <= 0) { 0 } else { y = 488.4 * l(x + 1000) - 3211 if (y < 0) 0 else y } scale = 0 (y + 0.5) / 1 " 2>/dev/null) # Fallback if bc fails (math domain error or missing utility) if ! [[ "$y" =~ ^[0-9]+$ ]]; then y=280 fi # Clamp values if (( y < MIN_TARGET_KBPS )); then y=$MIN_TARGET_KBPS elif (( y > MAX_TARGET_KBPS )); then y=$MAX_TARGET_KBPS fi echo "$y" } # Analyze audio volume to determine normalization gain. # Uses the global TARGET_VOL to calculate the necessary adjustment. # Avoids awk; uses grep, sed, sort, and bc. detect_loudness_gain() { local input_file="$1" local duration duration=$(ffprobe -v quiet -show_entries format=duration -of csv=p=0 "$input_file" 2>/dev/null) if [[ -z "$duration" ]]; then echo "0.0" return fi local raw="" local len=60 # Sample 1 minute chunks # Check if file is short (< 15 mins = 900 seconds) if (( $(echo "$duration < 900" | bc -l 2>/dev/null || echo 0) )); then echo " -> File is short (< 15m), analyzing full volume..." >&2 # Note: We must use || true on the pipeline because grep returns 1 if no match found, # which would crash the script due to set -e / pipefail. raw=$(ffmpeg -i "$input_file" -af "volumedetect" -map 0:a -f null /dev/null 2>&1 |& grep 'max_volume' || true) else echo " -> File is long, sampling volume at 5 points..." >&2 local sample_positions=(0 0.25 0.5 0.75) local i=1 local pos t for pos in "${sample_positions[@]}"; do t=$(printf "%.3f" "$(echo "$duration * $pos" | bc -l)") raw+=$(ffmpeg -ss "$t" -t "$len" -i "$input_file" -af "volumedetect" -map 0:a -f null /dev/null 2>&1 |& grep 'max_volume' || true) raw+=$'\n' printf " %d" $i >&2 ((i++)) done # Final sample at end t=$(printf "%.3f" "$(echo "$duration - $len" | bc -l)") if (( $(echo "$t < 0" | bc -l 2>/dev/null || echo 0) )); then t=0 fi raw+=$(ffmpeg -ss "$t" -t "$len" -i "$input_file" -af "volumedetect" -map 0:a -f null /dev/null 2>&1 |& grep 'max_volume' || true) printf " %d\n" $i >&2 fi if [[ -z "$raw" ]]; then echo "0.0" return fi # Find loudest section (max_volume closest to 0) # 1. Extract number (e.g., -12.3) # 2. Sort numerically (-20, -15, -5) # 3. Tail gets -5 (loudest) local max_db max_db=$(echo "$raw" | sed -n 's/.*max_volume: \(-[0-9.]*\) dB.*/\1/p' | sort -n | tail -n 1) if [[ -z "$max_db" ]] || ! [[ "$max_db" =~ ^-?[0-9]+(\.[0-9]+)?$ ]]; then echo "0.0" return fi # Calculate gain: Gain = Target - Current_Max # Example: Current=-10dB, Target=-20dB -> Gain = -10dB (Reduction) # Example: Current=-30dB, Target=-20dB -> Gain = +10dB (Amplification) local gain gain=$(bc -l <<< "${TARGET_VOL} - ${max_db}") printf "%.2f" "$gain" } # Detect black bars via cropdetect # Samples frames at intervals to find consistent crop area detect_crop_filter() { local file="$1" local duration duration=$(get_duration "$file") || return 1 local dur_sec=${duration%.*} if (( dur_sec < 3 )); then return fi local num_samples=8 local interval=$(( dur_sec / (num_samples + 1) )) (( interval = interval > 0 ? interval : 1 )) local min_x1=99999 min_y1=99999 max_x2=-1 max_y2=-1 local found_any=0 for ((i = 1; i <= num_samples; i++)); do local ss=$(( i * interval )) local crop_line # cropdetect=limit=24: Only crop if black level is below 24. crop_line=$(ffmpeg -nostdin -hide_banner -ss "$ss" -i "$file" -t 1 -vf "cropdetect=limit=24:round=2:reset=0" -f null - 2>&1 | grep "crop=" | tail -n1) || continue if [[ -n "$crop_line" ]]; then # Extract params: "crop=1920:1080:0:0" -> "1920:1080:0:0" local crop_params="${crop_line#*crop=}" local w h x y IFS=':' read -r w h x y <<< "$crop_params" [[ "$w" =~ ^[0-9]+$ ]] && [[ "$h" =~ ^[0-9]+$ ]] && [[ "$x" =~ ^[0-9]+$ ]] && [[ "$y" =~ ^[0-9]+$ ]] || continue local x1=$((x)) local y1=$((y)) local x2=$((x + w)) local y2=$((y + h)) # Accumulate min/max for bounding box (( min_x1 = x1 < min_x1 ? x1 : min_x1 )) (( min_y1 = y1 < min_y1 ? y1 : min_y1 )) (( max_x2 = x2 > max_x2 ? x2 : max_x2 )) (( max_y2 = y2 > max_y2 ? y2 : max_y2 )) found_any=1 fi done if (( !found_any )); then return; fi local final_w=$(( max_x2 - min_x1 )) local final_h=$(( max_y2 - min_y1 )) local orig_w orig_h orig_w=$(ffprobe -v error -select_streams v:0 -show_entries stream=width -of csv=p=0 "$file") orig_h=$(ffprobe -v error -select_streams v:0 -show_entries stream=height -of csv=p=0 "$file") # Only apply crop if it actually reduces dimensions if (( final_w < orig_w || final_h < orig_h )); then echo "crop=${final_w}:${final_h}:${min_x1}:${min_y1}" fi } # ======================================== # MAIN SCRIPT # ======================================== # --- Argument Parsing --- SCALE_FACTOR="" input_file="" if [[ $# -eq 2 ]]; then SCALE_FACTOR="$1" input_file="$2" elif [[ $# -eq 1 ]]; then input_file="$1" else echo "Usage: $0 [scale_factor] <video_file>" >&2 echo " scale_factor: Optional. Divides target bitrate by this number." >&2 echo " e.g. 2.0 creates a file 1/2 the original size." >&2 exit 1 fi # --- Validation --- [[ -f "$input_file" ]] || die "Input file not found: $input_file" [[ -r "$input_file" ]] || die "Input file not readable: $input_file" filename=$(basename "$input_file") outfile="${filename%.*}.mkv" outdir="out" origdir="orig" mkdir -p "$outdir" "$origdir" echo "=======================================" echo "ANALYZING: $filename" echo "=======================================" # STEP 1: Extract Input Video Bitrate input_bitrate_bps=$(get_effective_video_bitrate_bps "$input_file") input_bitrate_kbps=$(( input_bitrate_bps / 1000 )) echo " -> Input video bitrate: ${input_bitrate_kbps} kbps" # STEP 1.5: Skip Check if (( input_bitrate_kbps < 256 )); then echo " -> SKIP: Input bitrate (${input_bitrate_kbps} kbps) is below threshold (256 kbps)." echo " -> File considered already compressed or low quality." exit 0 fi # STEP 2: Compute Target Bitrate if [[ -n "$SCALE_FACTOR" ]]; then # Manual Scaling Mode # Validate scale_factor is a positive number if ! [[ "$SCALE_FACTOR" =~ ^[0-9]+(\.[0-9]+)?$ ]]; then die "Scale factor must be a positive number (e.g., 1, 1.5, 2)." fi target_bitrate_kbps=$(echo "scale=0; $input_bitrate_kbps / $SCALE_FACTOR" | bc) echo " -> MANUAL MODE: Scaling bitrate by 1/${SCALE_FACTOR}." else # Auto Curve Mode target_bitrate_kbps=$(calculate_target_bitrate_kbps "$input_bitrate_kbps") echo " -> AUTO MODE: Logarithmic targeting." fi target_bitrate_bps=$(( target_bitrate_kbps * 1000 )) echo " -> Target video bitrate: ${target_bitrate_kbps} kbps" # STEP 3: Estimate Output Size duration_sec=$(get_duration "$input_file") || die "Cannot determine duration" total_bitrate_kbps=$(( target_bitrate_kbps + 12 )) # +12k audio (approx) estimated_size_bytes=$(bc -l <<< "($total_bitrate_kbps * 1000 * $duration_sec) / 8" 2>/dev/null) estimated_size_bytes=${estimated_size_bytes%.*} # truncate float input_size_mb=$(bc -l <<< "$(stat -c%s "$input_file") / (1024*1024)" 2>/dev/null) input_size_mb=${input_size_mb%.*} estimated_size_mb=$(bc -l <<< "scale=1; $estimated_size_bytes / (1024*1024)" 2>/dev/null) [[ -z "$estimated_size_mb" ]] && estimated_size_mb="0.0" echo " -> Input file size: ${input_size_mb} MB" echo " -> Estimated output size: ${estimated_size_mb} MB" # STEP 4: Audio & Crop Analysis volume_gain=$(detect_loudness_gain "$input_file") echo " -> Audio gain for ${TARGET_VOL} dB target: ${volume_gain}dB" crop_filter="" if crop_result=$(detect_crop_filter "$input_file"); then crop_filter="$crop_result" fi if [[ -n "$crop_filter" ]]; then echo " -> Crop filter: $crop_filter" else echo " -> No significant black bars detected." fi # STEP 5: Build Filter Chain if [[ -n "$crop_filter" ]]; then video_filter="${crop_filter},${SCALE_FILTER},mpdecimate" else video_filter="$SCALE_FILTER" fi audio_filter="volume=${volume_gain}dB" passlogfile="${outdir}/${filename%.*}.log" echo "--------------------------------------------------" echo "ENCODING PLAN:" echo " Video filter: $video_filter" echo " Audio filter: $audio_filter" echo " Output: ${outdir}/${outfile}" echo "--------------------------------------------------" # Prompt user before proceeding (Destructive action) #read -r -p "Proceed with encoding and archiving original? (y/N): " confirm #if [[ ! "$confirm" =~ ^[Yy]$ ]]; then # echo "Aborted by user." # exit 0 #fi #-c:v "$VIDEO_CODEC" -b:v "${target_bitrate_bps}" \ # ---------------------------- # SINGLE-PASS HEVC ENCODING # ---------------------------- echo "ENCODE: Creating final output..." if ! ffmpeg -y -hide_banner -i "$input_file" \ -filter:v "$video_filter" \ -filter:a "$audio_filter" \ -preset "$ENCODING_PRESET" \ -fps_mode vfr \ -c:v "$VIDEO_CODEC" -crf "32" \ -g 600 -keyint_min 600 \ -x265-params "$ENCODING_PARAMS" \ -c:a libopus -b:a "$AUDIO_BITRATE" -application voip -frame_duration 20 -compression_level 10 \ -movflags +faststart -tune fastdecode -metadata handler_name="spumco" \ "${outdir}/${outfile}"; then echo "FAILURE: Encoding failed!" >&2 exit 1 fi # Cleanup logs rm -f "${passlogfile}"* # ---------------------------- # POST-PROCESSING # ---------------------------- echo "=======================================" echo "SUCCESS: Output written to ${outdir}/${outfile}" echo "Archiving original to ${origdir}/..." mv "$input_file" "$origdir/" printf '\a'; sleep 0.5; printf '\a'; sleep 0.5; printf '\a' echo "Done." exit 0