Agent Skills: FFmpeg Video Analysis Filters (2025)

CRITICAL GUIDELINES

Windows File Path Requirements

MANDATORY: Always Use Backslashes on Windows for File Paths

When using Edit or Write tools on Windows, you MUST use backslashes (\) in file paths, NOT forward slashes (/).

Quick Reference

| Task | Filter | Command Pattern | |------|--------|-----------------| | Detect black frames | blackdetect | -vf blackdetect=d=0.5:pic_th=0.98 | | Detect frozen frames | freezedetect | -vf freezedetect=n=0.003:d=2 | | Detect blur | blurdetect | -vf blurdetect=low=5:high=15 | | Auto crop | cropdetect | -vf cropdetect=24:16:0 | | Scene changes | scdet | -vf scdet=threshold=10 | | Quality metrics | psnr, ssim | -lavfi "[0:v][1:v]psnr" -f null - | | Frame info | showinfo | -vf showinfo |

When to Use This Skill

Use for quality control and automation workflows:

• Automated video analysis pipelines
• Detecting problematic frames (black, frozen, blurry)
• Finding optimal crop parameters
• Measuring quality after encoding
• Broadcast compliance checking
• Content-aware editing decisions

FFmpeg Video Analysis Filters (2025)

Comprehensive guide to video analysis filters for quality control, automation, and professional workflows.

Detection Filters

blackdetect - Detect Black Frames

Detects video sequences that are completely black, useful for finding commercial breaks, scene boundaries, or encoding issues.

# Basic black detection
ffmpeg -i input.mp4 -vf "blackdetect=d=0.5:pic_th=0.98:pix_th=0.10" -f null -

# More sensitive detection (darker threshold)
ffmpeg -i input.mp4 -vf "blackdetect=d=0.1:pic_th=0.90" -f null -

# Save detection results to log
ffmpeg -i input.mp4 -vf "blackdetect=d=0.5" -f null - 2>&1 | grep blackdetect

# Detect and extract black segments
ffmpeg -i input.mp4 -vf "blackdetect=d=2.0" -f null - 2>&1 | \
  grep -oP 'black_start:\K[0-9.]+|black_end:\K[0-9.]+'

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | d | Minimum duration (seconds) | 2.0 | > 0 | | pic_th | Picture black ratio threshold | 0.98 | 0-1 | | pix_th | Pixel black threshold | 0.10 | 0-1 |

Output format:

[blackdetect @ 0x...] black_start:10.5 black_end:12.3 black_duration:1.8

blackframe - Detect Nearly Black Frames

Similar to blackdetect but reports individual frames and their blackness amount.

# Detect frames that are 98% black
ffmpeg -i input.mp4 -vf "blackframe=amount=98:threshold=32" -f null -

# Output includes frame number and percentage
ffmpeg -i input.mp4 -vf "blackframe=amount=90" -f null - 2>&1 | grep blackframe

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | amount | Percentage threshold | 98 | 0-100 | | threshold | Pixel brightness threshold | 32 | 0-255 |

freezedetect - Detect Frozen Frames

Detects sequences where the video appears frozen (repeated frames).

# Basic freeze detection
ffmpeg -i input.mp4 -vf "freezedetect=n=0.003:d=2" -f null -

# More sensitive (detect shorter freezes)
ffmpeg -i input.mp4 -vf "freezedetect=n=0.001:d=0.5" -f null -

# Very strict (only perfectly identical frames)
ffmpeg -i input.mp4 -vf "freezedetect=n=0:d=1" -f null -

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | n | Noise tolerance (frame diff) | 0.001 | 0-1 | | d | Minimum freeze duration | 2.0 | > 0 |

Output format:

[freezedetect @ 0x...] freeze_start: 45.2
[freezedetect @ 0x...] freeze_duration: 3.5
[freezedetect @ 0x...] freeze_end: 48.7

blurdetect - Detect Blurry Frames

Detects frames that are out of focus or motion blurred.

# Basic blur detection
ffmpeg -i input.mp4 -vf "blurdetect=low=5:high=15:radius=50" -f null -

# More sensitive detection
ffmpeg -i input.mp4 -vf "blurdetect=low=3:high=10" -f null -

# Output blur values per frame
ffmpeg -i input.mp4 -vf "blurdetect,metadata=print:file=blur.txt" -f null -

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | low | Low edge threshold | 5 | 1-100 | | high | High edge threshold | 15 | 1-100 | | radius | Search radius | 50 | 1-100 | | block_pct | Block percentage | 80 | 0-100 | | block_width | Block width | - | > 0 | | block_height | Block height | - | > 0 | | planes | Planes to analyze | 1 | 0-15 |

Output metadata:

• lavfi.blur - Blur value (lower = blurrier)

scdet - Scene Change Detection

Detects scene changes based on frame-to-frame differences.

# Basic scene detection
ffmpeg -i input.mp4 -vf "scdet=threshold=10:sc_pass=1" -f null -

# More sensitive (detect more scene changes)
ffmpeg -i input.mp4 -vf "scdet=threshold=5" -f null -

# Output scene changes with timestamps
ffmpeg -i input.mp4 -vf "scdet=t=10,metadata=print:file=scenes.txt" -f null -

# Combined with select filter to extract scene thumbnails
ffmpeg -i input.mp4 -vf "scdet=threshold=10,select='gt(scene,0.4)',showinfo" \
  -vsync vfr scene_%04d.jpg

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | threshold / t | Scene change threshold | 10.0 | 0-100 | | sc_pass | Pass scene score to output | 0 | 0-1 |

Output metadata:

• lavfi.scd.score - Scene change score (0-1)
• lavfi.scd.mafd - Mean absolute frame difference
• lavfi.scd.time - Timestamp of scene change

cropdetect - Auto Crop Detection

Automatically detects optimal crop values to remove black borders.

# Basic crop detection
ffmpeg -i input.mp4 -vf "cropdetect=24:16:0" -f null -

# More aggressive detection (lower threshold)
ffmpeg -i input.mp4 -vf "cropdetect=16:2:0" -f null -

# Detect and apply crop in one command
crop=$(ffmpeg -i input.mp4 -vf "cropdetect=24:16:0" -f null - 2>&1 | \
  grep -oP 'crop=\K[0-9:]+' | tail -1)
ffmpeg -i input.mp4 -vf "crop=$crop" output.mp4

# Detect letterbox dimensions
ffmpeg -i input.mp4 -vf "cropdetect=round=2:reset=0" -f null - 2>&1 | grep crop

Parameters: | Parameter | Description | Default | Range | |-----------|-------------|---------|-------| | limit | Threshold for black pixels | 24 | 0-255 | | round | Round to nearest multiple | 16 | >= 2 | | reset | Reset counter (frames) | 0 | >= 0 | | skip | Skip initial frames | 0 | >= 0 |

Output format:

[cropdetect @ 0x...] x1:0 x2:1919 y1:140 y2:939 w:1920 h:800 x:0 y:140 crop=1920:800:0:140

idet - Interlace Detection

Detects whether video is interlaced and identifies field order.

# Basic interlace detection
ffmpeg -i input.mp4 -vf "idet" -frames:v 500 -f null -

# Output detection summary
ffmpeg -i input.mp4 -vf "idet" -f null - 2>&1 | grep -A5 "Repeated Fields"

Output includes:

• Single (progressive frames)
• Multi (interlaced - multiple fields from same frame)
• Repeated (repeated fields - pulldown)
• Top Field First (TFF) vs Bottom Field First (BFF)

Quality Metrics

psnr - Peak Signal-to-Noise Ratio

Compares two videos and outputs PSNR quality metric.

# Compare original vs encoded
ffmpeg -i original.mp4 -i encoded.mp4 \
  -lavfi "[0:v][1:v]psnr" -f null -

# Output PSNR per frame to file
ffmpeg -i original.mp4 -i encoded.mp4 \
  -lavfi "[0:v][1:v]psnr=stats_file=psnr.log" -f null -

# Get average PSNR only
ffmpeg -i original.mp4 -i encoded.mp4 \
  -lavfi "[0:v][1:v]psnr" -f null - 2>&1 | grep "average"

Output format:

[Parsed_psnr_0 @ 0x...] PSNR y:45.123 u:48.456 v:49.789 average:46.234 min:35.123 max:inf

Quality guidelines: | PSNR (dB) | Quality | |-----------|---------| | > 40 | Excellent (indistinguishable) | | 35-40 | Good | | 30-35 | Fair | | < 30 | Poor |

ssim - Structural Similarity Index

More perceptually accurate than PSNR for quality comparison.

# Compare original vs encoded
ffmpeg -i original.mp4 -i encoded.mp4 \
  -lavfi "[0:v][1:v]ssim" -f null -

# Output SSIM per frame to file
ffmpeg -i original.mp4 -i encoded.mp4 \
  -lavfi "[0:v][1:v]ssim=stats_file=ssim.log" -f null -

Output format:

[Parsed_ssim_0 @ 0x...] SSIM Y:0.987 (18.87) U:0.992 (20.97) V:0.993 (21.55) All:0.989 (19.59)

Quality guidelines: | SSIM | Quality | |------|---------| | > 0.98 | Excellent | | 0.95-0.98 | Good | | 0.90-0.95 | Fair | | < 0.90 | Poor |

vmafmotion - VMAF Motion Score

Calculates motion activity score used by VMAF.

# Calculate motion score
ffmpeg -i input.mp4 -vf "vmafmotion" -f null - 2>&1 | grep vmafmotion

# Get average motion
ffmpeg -i input.mp4 -vf "vmafmotion" -f null - 2>&1 | tail -1

signalstats - Video Signal Statistics

Comprehensive signal analysis for broadcast QC.

# Full signal analysis
ffmpeg -i input.mp4 -vf "signalstats=stat=tout+vrep+brng" -f null -

# Output to file
ffmpeg -i input.mp4 -vf "signalstats,metadata=print:file=stats.txt" -f null -

# Check for broadcast-safe levels
ffmpeg -i input.mp4 -vf "signalstats=stat=brng,metadata=print" -f null - 2>&1 | \
  grep "lavfi.signalstats.BRNG"

Statistics available: | Stat | Description | |------|-------------| | tout | Temporal outliers | | vrep | Vertical line repetition | | brng | Broadcast range violations |

Output metadata includes:

• YMIN, YMAX - Luma range
• YAVG - Average luma
• UMIN, UMAX, VMIN, VMAX - Chroma range
• SATMIN, SATMAX, SATAVG - Saturation
• HUEAVG - Average hue
• BRNG - Out of broadcast range pixel count

Frame Information

showinfo - Display Frame Information

Outputs detailed information about each frame.

# Show all frame info
ffmpeg -i input.mp4 -vf "showinfo" -f null -

# Show specific frames only
ffmpeg -i input.mp4 -vf "select='eq(n,0)+eq(n,100)',showinfo" -f null -

# Parse specific information
ffmpeg -i input.mp4 -vf "showinfo" -f null - 2>&1 | grep "pts_time"

Output includes:

[Parsed_showinfo_0 @ 0x...] n:   0 pts:      0 pts_time:0       duration:   1001
 duration_time:0.0417083 fmt:yuv420p cl:left sar:1/1 s:1920x1080 i:P iskey:1
 type:I checksum:12345678 plane_checksum:[AAAAAAAA BBBBBBBB CCCCCCCC]

Fields:

• n - Frame number
• pts - Presentation timestamp
• pts_time - PTS in seconds
• duration - Frame duration
• fmt - Pixel format
• s - Size (resolution)
• i - Interlaced (P=progressive, T=top, B=bottom)
• iskey - Is keyframe
• type - Frame type (I/P/B)

siti - Spatial and Temporal Information

ITU-T P.910 compliant SI/TI calculation.

# Calculate SI/TI
ffmpeg -i input.mp4 -vf "siti" -f null - 2>&1 | grep siti

# Output to file
ffmpeg -i input.mp4 -vf "siti=print_summary=1" -f null -

Output:

• SI - Spatial Information (edge/texture complexity)
• TI - Temporal Information (motion activity)

Automation Patterns

QC Pipeline Script

#!/bin/bash
# Comprehensive video QC analysis

INPUT="$1"
OUTPUT_DIR="qc_results"
mkdir -p "$OUTPUT_DIR"

echo "Analyzing: $INPUT"

# 1. Black frame detection
ffmpeg -i "$INPUT" -vf "blackdetect=d=0.5" -f null - 2>&1 | \
  grep blackdetect > "$OUTPUT_DIR/black_frames.txt"

# 2. Freeze detection
ffmpeg -i "$INPUT" -vf "freezedetect=n=0.003:d=2" -f null - 2>&1 | \
  grep freeze > "$OUTPUT_DIR/frozen_frames.txt"

# 3. Scene detection
ffmpeg -i "$INPUT" -vf "scdet=threshold=10,metadata=print:file=$OUTPUT_DIR/scenes.txt" -f null -

# 4. Crop detection
ffmpeg -i "$INPUT" -vf "cropdetect=24:16:0" -f null - 2>&1 | \
  grep crop | tail -1 > "$OUTPUT_DIR/crop.txt"

# 5. Interlace detection
ffmpeg -i "$INPUT" -vf "idet" -frames:v 500 -f null - 2>&1 | \
  grep -A5 "Repeated" > "$OUTPUT_DIR/interlace.txt"

# 6. Signal stats
ffmpeg -i "$INPUT" -vf "signalstats=stat=brng,metadata=print:file=$OUTPUT_DIR/signal.txt" \
  -f null -

echo "QC analysis complete. Results in $OUTPUT_DIR/"

Extract Frames at Scene Changes

# Extract one frame per scene
ffmpeg -i input.mp4 \
  -vf "scdet=threshold=10,select='gt(scene,0.4)'" \
  -vsync vfr \
  -frame_pts 1 \
  scene_%04d.jpg

# Extract with timestamp in filename
ffmpeg -i input.mp4 \
  -vf "scdet=threshold=10,select='gt(scene,0.4)',showinfo" \
  -vsync vfr \
  "scene_%04d_%{pts}.jpg"

Quality Comparison Batch

#!/bin/bash
# Compare multiple encodes against reference

REFERENCE="reference.mp4"

for encoded in encode_*.mp4; do
  echo "Comparing: $encoded"

  # PSNR
  psnr=$(ffmpeg -i "$REFERENCE" -i "$encoded" \
    -lavfi "[0:v][1:v]psnr" -f null - 2>&1 | \
    grep -oP 'average:\K[0-9.]+')

  # SSIM
  ssim=$(ffmpeg -i "$REFERENCE" -i "$encoded" \
    -lavfi "[0:v][1:v]ssim" -f null - 2>&1 | \
    grep -oP 'All:\K[0-9.]+')

  echo "  PSNR: $psnr dB, SSIM: $ssim"
done

Best Practices

1. Use appropriate thresholds - Start with defaults, adjust based on content
2. Sample long videos - Use -t to analyze portions first
3. Combine filters - Chain detection filters for comprehensive analysis
4. Parse output - Use grep/awk to extract relevant data
5. Batch processing - Create scripts for consistent QC workflows
6. Log results - Use metadata=print:file= to save results

This guide covers video analysis filters for 2025. For encoding quality, see ffmpeg-fundamentals-2025. For hardware-accelerated analysis, check GPU filter support in ffmpeg-hardware-acceleration.