xiaohongshu
xiaohongshu
ffmpeg-noise-reduction
Complete FFmpeg noise reduction and denoising for video and audio. PROACTIVELY activate for: (1) Video denoising (nlmeans, hqdn3d, vaguedenoiser), (2) Hardware-accelerated denoising (nlmeans_opencl, nlmeans_vulkan), (3) Audio noise reduction (afftdn, anlmdn), (4) Film grain removal, (5) Low-light...
Full skill instructions
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 (/).
| Task | Filter | Command Pattern |
|---|---|---|
| Video denoising (best) | nlmeans | -vf nlmeans=s=3:p=7:r=15 |
| Video denoising (fast) | hqdn3d | -vf hqdn3d=4:3:6:4.5 |
| GPU video denoising | nlmeans_vulkan | -vf nlmeans_vulkan |
| Audio denoising | afftdn | -af afftdn=nf=-25 |
| Adaptive temporal | atadenoise | -vf atadenoise |
Use for noise reduction workflows:
Comprehensive guide to video and audio denoising filters.
The highest quality software denoiser, using non-local means algorithm.
# Basic denoising
ffmpeg -i noisy.mp4 -vf "nlmeans" output.mp4
# Medium denoising (good default)
ffmpeg -i noisy.mp4 -vf "nlmeans=s=3.0:p=7:r=15" output.mp4
# Strong denoising (very noisy footage)
ffmpeg -i noisy.mp4 -vf "nlmeans=s=5.0:p=7:r=15" output.mp4
# Light denoising (preserve detail)
ffmpeg -i noisy.mp4 -vf "nlmeans=s=1.5:p=5:r=9" output.mp4
# Denoise only chroma (preserve luma detail)
ffmpeg -i noisy.mp4 -vf "nlmeans=s=3.0:sc=5.0:p=7:r=15" output.mp4
Parameters:
| Parameter | Description | Default | Range |
|---|---|---|---|
s | Denoising strength (sigma) | 1.0 | 0-30 |
p | Patch size | 7 | 1-99 (odd) |
pc | Patch size for chroma | 0 (same as p) | 1-99 |
r | Research window size | 15 | 1-99 (odd) |
rc | Research window for chroma | 0 (same as r) | 1-99 |
Strength guidelines:
| s value | Use case |
|---|---|
| 1.0-2.0 | Light noise, preserve detail |
| 2.0-4.0 | Moderate noise (typical) |
| 4.0-6.0 | Heavy noise, low light |
| 6.0+ | Extreme noise (quality loss) |
OpenCL accelerated version, much faster.
# OpenCL denoising
ffmpeg -i noisy.mp4 \
-vf "hwupload,nlmeans_opencl=s=3:p=7:r=15,hwdownload,format=yuv420p" \
output.mp4
# Strong denoising with OpenCL
ffmpeg -i noisy.mp4 \
-vf "hwupload,nlmeans_opencl=s=5:p=7:r=21,hwdownload,format=yuv420p" \
output.mp4
Cross-platform GPU acceleration using Vulkan.
# Vulkan denoising
ffmpeg -init_hw_device vulkan \
-i noisy.mp4 \
-vf "hwupload,nlmeans_vulkan=s=3:p=7:r=15,hwdownload,format=yuv420p" \
output.mp4
# Full Vulkan pipeline
ffmpeg -init_hw_device vulkan=vk \
-filter_hw_device vk \
-hwaccel vulkan -hwaccel_output_format vulkan \
-i noisy.mp4 \
-vf "nlmeans_vulkan=s=3" \
-c:v h264_vulkan output.mp4
Faster than nlmeans with good results for moderate noise.
# Default settings
ffmpeg -i noisy.mp4 -vf "hqdn3d" output.mp4
# Custom settings (luma_spatial:chroma_spatial:luma_tmp:chroma_tmp)
ffmpeg -i noisy.mp4 -vf "hqdn3d=4:3:6:4.5" output.mp4
# Strong spatial denoising
ffmpeg -i noisy.mp4 -vf "hqdn3d=8:6:6:4" output.mp4
# Strong temporal denoising (for static scenes)
ffmpeg -i noisy.mp4 -vf "hqdn3d=4:3:12:9" output.mp4
# Light denoising
ffmpeg -i noisy.mp4 -vf "hqdn3d=2:1.5:3:2" output.mp4
Parameters:
| Parameter | Description | Default | Range |
|---|---|---|---|
luma_spatial | Luma spatial strength | 4.0 | 0-255 |
chroma_spatial | Chroma spatial strength | 3.0 | 0-255 |
luma_tmp | Luma temporal strength | 6.0 | 0-255 |
chroma_tmp | Chroma temporal strength | 4.5 | 0-255 |
Temporal denoiser that preserves motion.
# Basic temporal denoising
ffmpeg -i noisy.mp4 -vf "atadenoise" output.mp4
# Stronger denoising
ffmpeg -i noisy.mp4 -vf "atadenoise=0a=0.1:0b=0.2:1a=0.1:1b=0.2:2a=0.1:2b=0.2" output.mp4
# Adjust threshold and planes
ffmpeg -i noisy.mp4 -vf "atadenoise=s=9:p=5" output.mp4
Parameters:
| Parameter | Description | Default |
|---|---|---|
0a, 1a, 2a | Threshold A for planes 0,1,2 | 0.02 |
0b, 1b, 2b | Threshold B for planes 0,1,2 | 0.04 |
s | Number of frames to average | 9 |
p | Planes to process | 7 (all) |
Uses wavelet transform for denoising.
# Basic wavelet denoising
ffmpeg -i noisy.mp4 -vf "vaguedenoiser" output.mp4
# Adjust threshold
ffmpeg -i noisy.mp4 -vf "vaguedenoiser=threshold=3:method=soft" output.mp4
# Stronger denoising
ffmpeg -i noisy.mp4 -vf "vaguedenoiser=threshold=6:nsteps=6" output.mp4
Parameters:
| Parameter | Description | Default | Values |
|---|---|---|---|
threshold | Denoising threshold | 2 | 0-inf |
method | Thresholding method | hard | hard, soft, garrote |
nsteps | Decomposition steps | 6 | 1-32 |
percent | Percent of full denoising | 85 | 0-100 |
Frequency domain denoising.
# Basic FFT denoising
ffmpeg -i noisy.mp4 -vf "fftdnoiz" output.mp4
# Adjust sigma
ffmpeg -i noisy.mp4 -vf "fftdnoiz=sigma=8" output.mp4
# Strong denoising
ffmpeg -i noisy.mp4 -vf "fftdnoiz=sigma=15:block=32:overlap=0.8" output.mp4
Another wavelet-based option.
# Basic wavelet denoising
ffmpeg -i noisy.mp4 -vf "owdenoise" output.mp4
# Adjust depth and strength
ffmpeg -i noisy.mp4 -vf "owdenoise=depth=10:ls=3:cs=3" output.mp4
Classic grain/noise removal (from AviSynth).
# Mode 1 (basic)
ffmpeg -i noisy.mp4 -vf "removegrain=m0=1" output.mp4
# Different modes for different noise types
ffmpeg -i noisy.mp4 -vf "removegrain=m0=17:m1=17:m2=17" output.mp4
Powerful audio noise reduction using FFT.
# Basic audio denoising
ffmpeg -i noisy_audio.mp4 -af "afftdn" output.mp4
# Specify noise floor
ffmpeg -i noisy_audio.mp4 -af "afftdn=nf=-25" output.mp4
# Stronger noise reduction
ffmpeg -i noisy_audio.mp4 -af "afftdn=nf=-20:nt=w" output.mp4
# With noise profile (sample noise-only section first)
ffmpeg -i noisy_audio.mp4 -af "afftdn=nr=10:nf=-30:nt=w:om=o" output.mp4
# Adaptive noise floor
ffmpeg -i noisy_audio.mp4 -af "afftdn=nt=c:om=o:bn=9" output.mp4
Parameters:
| Parameter | Description | Default | Range |
|---|---|---|---|
nr | Noise reduction (dB) | 12 | 0.01-97 |
nf | Noise floor (dB) | -50 | -80 to -20 |
nt | Noise type | w | w=white, v=vinyl, s=shellac, c=custom |
bn | Number of noise bands | 15 | 3-33 |
rf | Residual floor (dB) | -38 | -80 to -20 |
om | Output mode | o | i=input, o=output, n=noise |
Audio equivalent of video nlmeans.
# Basic NLM audio denoising
ffmpeg -i noisy_audio.mp4 -af "anlmdn" output.mp4
# Adjust strength
ffmpeg -i noisy_audio.mp4 -af "anlmdn=s=0.0001" output.mp4
# Stronger denoising
ffmpeg -i noisy_audio.mp4 -af "anlmdn=s=0.001:p=0.01" output.mp4
Parameters:
| Parameter | Description | Default |
|---|---|---|
s | Denoising strength | 0.00001 |
p | Patch radius | 0.002 |
r | Research radius | 0.006 |
m | Output mode | o (output) |
Removes impulsive noise from audio.
# Remove clicks
ffmpeg -i clicked_audio.mp4 -af "adeclick" output.mp4
# Adjust parameters
ffmpeg -i clicked_audio.mp4 -af "adeclick=w=55:o=75" output.mp4
Repairs clipped audio.
# Fix clipping
ffmpeg -i clipped_audio.mp4 -af "adeclip" output.mp4
# Denoise + brighten + contrast
ffmpeg -i dark_noisy.mp4 \
-vf "nlmeans=s=4:p=7:r=15,eq=brightness=0.1:contrast=1.2:saturation=1.1" \
-c:v libx264 -crf 18 \
enhanced.mp4
# Clean both video and audio noise
ffmpeg -i interview.mp4 \
-vf "hqdn3d=3:2:4:3" \
-af "afftdn=nf=-30:nr=10,loudnorm" \
-c:v libx264 -crf 20 \
-c:a aac -b:a 192k \
interview_clean.mp4
# Strong nlmeans for film grain
ffmpeg -i grainy_film.mp4 \
-vf "nlmeans=s=5:p=7:r=21" \
-c:v libx264 -crf 18 -preset slow \
degrained.mp4
# Preserve some grain (artistic choice)
ffmpeg -i grainy_film.mp4 \
-vf "nlmeans=s=2.5:p=5:r=11" \
-c:v libx264 -crf 18 \
light_degrain.mp4
# Remove blocking artifacts from highly compressed video
ffmpeg -i compressed.mp4 \
-vf "hqdn3d=2:1.5:3:2,unsharp=5:5:0.3:5:5:0" \
-c:v libx264 -crf 18 \
cleaner.mp4
#!/bin/bash
# Process multiple noisy videos with GPU
for video in noisy_*.mp4; do
output="clean_${video}"
ffmpeg -init_hw_device vulkan \
-i "$video" \
-vf "hwupload,nlmeans_vulkan=s=3,hwdownload,format=yuv420p" \
-c:v libx264 -crf 18 \
"$output"
done
| Filter | Quality | Speed | GPU Option | Best For |
|---|---|---|---|---|
nlmeans | Excellent | Slow | Yes (OpenCL/Vulkan) | Final output |
hqdn3d | Good | Fast | No | Real-time, previews |
atadenoise | Good | Medium | No | Static scenes |
vaguedenoiser | Good | Medium | No | Mixed content |
fftdnoiz | Good | Medium | No | Periodic noise |
| Filter | Quality | Speed | Best For |
|---|---|---|---|
afftdn | Excellent | Fast | All-purpose |
anlmdn | Good | Medium | Speech |
adeclick | Specialized | Fast | Clicks/pops |
This guide covers noise reduction for 2025. For video analysis and detection, see ffmpeg-video-analysis. For hardware acceleration, see ffmpeg-hardware-acceleration.
xiaohongshu
technical spec
product ux expert
database patterns
Conduct multi-agent task orchestration and workflow coordination.
Initialize project with Conductor artifacts (product definition,
Expert in web animations, transitions, and motion design using Framer Motion and CSS
Creates Mermaid and ASCII diagrams for flowcharts, architecture, ERDs, state machines, mindmaps, and more. Use when user mentions diagram, flowchart, mermaid, ASCII diagram, text diagram, terminal diagram, visualize, C4, mindmap, architecture diagram, sequence diagram, ERD, or needs visual docume...
PostgreSQL bindings for H3 hexagonal grid system. Use when working with H3 cells in Postgres, including spatial indexing, geometry/geography integration, and raster analysis.
Context-Driven Development skill for projects using Conductor. Use this skill when you detect a `conductor/` directory in the project, when working on tasks defined in a `plan.md` file, or when the user asks about tracks, specs, or plans. Automatically applies TDD workflow, tracks task completion...
Display project status, active tracks, and next actions
Official Stakpak application containerization standard operating procedure, a step-by-step guidline to properly dockerize applications. This is a rule book curated by the Stakpak Team.
Generate, edit, and beat-sync AI video with leading models in one workspace.
The world's fastest calendar for remote work
Transform Your Design with AI Designer by ImgCreator.ai
Revolutionizing Video Production with AI-Powered Creativity
Extend an image past the frame and let AI fill the new aspect ratio.
Discover your celebrity doppelgänger with StarByFace!
ChainClarity explains 700+ crypto whitepapers in plain English, with layered summaries, comparisons, research tools, alerts, and a $4.99 Pro plan.
Opus.ai: Revolutionize Your Web Experience