JosiahSiegel/ffmpeg-hardware-acceleration
plugins/ffmpeg-core/skills/ffmpeg-hardware-acceleration/SKILL.md
Complete GPU-accelerated encoding/decoding system for FFmpeg 7.1 LTS and 8.0.1 (latest stable, released 2025-11-20). PROACTIVELY activate for: (1) NVIDIA NVENC/NVDEC encoding, (2) Intel Quick Sync Video (QSV), (3) AMD AMF encoding, (4) Apple VideoToolbox, (5) Linux VAAPI setup, (6) Vulkan Video 8.0 (FFv1, AV1, VP9, ProRes RAW), (7) VVC/H.266 hardware decoding (VAAPI/QSV), (8) GPU pipeline optimization with pad_cuda, (9) Docker GPU containers, (10) Performance benchmarking. Provides: Platform-specific commands, preset comparisons, quality tuning, full GPU pipeline examples, Vulkan compute codecs, VVC decoding, troubleshooting guides. Ensures: Maximum encoding speed with optimal quality using GPU acceleration.
$ install --global
skillsauthnpx skillsauth add JosiahSiegel/claude-plugin-marketplace ffmpeg-hardware-accelerationInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
Security Scan Results
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
3
Scanners Passed
9
Scanners in report
Clean
TrivyContainer and dependency vulnerability scanner
95%
Clean
SemgrepStatic code analysis for vulnerabilities
95%
Clean
mcp-scan (Snyk)Model Context Protocol security validation
95%
Skipped
Snyk (dep)Open source security scanning
50%
Skipped
Socket.devSupply chain security analysis
50%
Skipped
VirusTotalMulti-engine malware detection
50%
Skipped
CrowdStrikeAdvanced threat intelligence
50%
Skipped
OSV-ScannerOpen Source Vulnerability database check
50%
Skipped
OWASP Dep-Check
50%
Last scanned: May 21, 2026, 8:03 AM109.6s7 files scanned
SKILL.md
- name:
- ffmpeg-hardware-acceleration
- description:
- |
- PROACTIVELY activate for:
- (1) NVIDIA NVENC/NVDEC encoding, (2) Intel Quick Sync Video (QSV), (3) AMD AMF encoding, (4) Apple VideoToolbox, (5) Linux VAAPI setup, (6) Vulkan Video 8.0 (FFv1, AV1, VP9, ProRes RAW), (7) VVC/H.266 hardware decoding (VAAPI/QSV), (8) GPU pipeline optimization with pad_cuda, (9) Docker GPU containers, (10) Performance benchmarking.
- Provides:
- Platform-specific commands, preset comparisons, quality tuning, full GPU pipeline examples, Vulkan compute codecs, VVC decoding, troubleshooting guides.
- Ensures:
- Maximum encoding speed with optimal quality using GPU acceleration.
When to Use This Skill
Activate when GPU acceleration is needed:
- Encoding speed is critical (10-30x faster than CPU)
- Processing large batches of videos
- Real-time encoding for streaming
- Server-side transcoding at scale
- Docker containers with GPU passthrough
GPU encoding trades some quality for massive speed. Use -cq, -qp, or -global_quality for quality control.
Quick Reference
| Platform | Encoder | Decoder | Detect Command |
|----------|---------|---------|----------------|
| NVIDIA | h264_nvenc, hevc_nvenc, av1_nvenc | h264_cuvid, hevc_cuvid | ffmpeg -encoders \| grep nvenc |
| Intel QSV | h264_qsv, hevc_qsv, av1_qsv | h264_qsv, hevc_qsv | ffmpeg -encoders \| grep qsv |
| AMD AMF | h264_amf, hevc_amf, av1_amf | N/A (use software) | ffmpeg -encoders \| grep amf |
| Apple | h264_videotoolbox, hevc_videotoolbox | h264_videotoolbox | macOS only |
| VAAPI | h264_vaapi, hevc_vaapi, av1_vaapi | with -hwaccel vaapi | Linux only |
| Vulkan | h264_vulkan, hevc_vulkan, av1_vulkan, ffv1_vulkan | VP9, ProRes RAW (8.0+) | ffmpeg -encoders \| grep vulkan |
Current Latest: FFmpeg 8.0.1 (released 2025-11-20). Check with ffmpeg -version.
Hardware Acceleration Overview
Hardware acceleration uses dedicated GPU/SoC components for video processing:
- NVENC/NVDEC (NVIDIA): dedicated encode/decode engines
- QSV (Intel): Quick Sync Video on Intel CPUs with integrated graphics
- AMF (AMD): Advanced Media Framework for AMD GPUs
- VideoToolbox (Apple): macOS/iOS hardware acceleration
- VAAPI (Linux): Video Acceleration API (Intel, AMD on Linux)
- Vulkan Video (Cross-platform, FFmpeg 7.1+/8.0): compute-shader-based codecs
Performance Comparison (2025 Benchmarks)
| Method | Speed | Quality | Power | Use Case | |--------|-------|---------|-------|----------| | libx264 (CPU) | 1x | Best | High | Quality-critical | | libx265 (CPU) | 0.3x | Best | Very High | Archival | | h264_nvenc | 10-20x | Good | Low | Real-time, streaming | | hevc_nvenc | 8-15x | Good | Low | 4K streaming | | h264_qsv | 8-15x | Good | Very Low | Laptop, efficiency | | h264_amf | 8-15x | Good | Low | AMD systems |
Core Workflow
- Detect what you have:
ffmpeg -hwaccels,ffmpeg -encoders | grep <api> - Pick a backend: prefer the vendor-native one (NVENC on NVIDIA, QSV on Intel, AMF on AMD, VideoToolbox on Apple). Use Vulkan for cross-platform portability.
- Build a full-GPU pipeline where possible:
- Add
-hwaccel <api>and-hwaccel_output_format <api>before-i - Use GPU-side filters (
scale_cuda,scale_vulkan,vpp_qsv, ...) - Use the matching hardware encoder
- Add
- Tune quality:
-preset(NVENCp1-p7),-cq/-qp/-global_quality, lookahead, spatial/temporal AQ - Verify: benchmark with
ffmpeg -benchmarkand monitor GPU vianvidia-smi dmon,intel_gpu_top, etc.
Minimal Examples per Backend
# NVIDIA NVENC
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input.mp4 \
-c:v h264_nvenc -preset p4 -b:v 5M output.mp4
# Intel QSV
ffmpeg -hwaccel qsv -hwaccel_output_format qsv -i input.mp4 \
-c:v h264_qsv -preset medium -b:v 5M output.mp4
# AMD AMF
ffmpeg -i input.mp4 -c:v h264_amf -quality balanced -b:v 5M output.mp4
# Apple VideoToolbox
ffmpeg -i input.mp4 -c:v h264_videotoolbox -b:v 5M output.mp4
# Linux VAAPI
ffmpeg -hwaccel vaapi -hwaccel_device /dev/dri/renderD128 \
-hwaccel_output_format vaapi -i input.mp4 \
-c:v h264_vaapi -b:v 5M output.mp4
# Vulkan (cross-platform)
ffmpeg -init_hw_device vulkan -i input.mp4 \
-c:v h264_vulkan -b:v 5M output.mp4
Full GPU Pipeline Pattern (Critical)
ffmpeg -y -vsync 0 \
-hwaccel cuda -hwaccel_output_format cuda \
-i input.mp4 \
-vf scale_cuda=1280:720 \
-c:v h264_nvenc -preset p4 -b:v 5M \
-c:a copy \
output.mp4
Omitting -hwaccel_output_format can cut throughput by up to 50% because decoded frames silently round-trip through CPU memory. See references/gpu-memory-and-troubleshooting.md for memory flow diagrams and best practices.
Cross-Vendor Filter Quick Map
| Operation | NVIDIA | Intel | AMD/Linux | Cross-platform |
|-----------|--------|-------|-----------|----------------|
| Scale | scale_cuda, scale_npp | vpp_qsv, scale_qsv | scale_vaapi | scale_vulkan, scale_opencl, libplacebo |
| Overlay | overlay_cuda | - | - | overlay_vulkan, overlay_opencl |
| Deinterlace | bwdif_cuda | vpp_qsv | deinterlace_vaapi | bwdif_vulkan |
| Denoise | bilateral_cuda | - | - | nlmeans_vulkan, nlmeans_opencl |
| Chromakey | chromakey_cuda | - | - | colorkey_opencl |
| Tonemap (HDR->SDR) | - | - | tonemap_vaapi | libplacebo, tonemap_opencl |
| Pad/letterbox | pad_cuda (8.0+) | - | - | pad_opencl |
Use-Case Quick Recipes
Live streaming (low latency)
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input \
-c:v h264_nvenc -preset p3 -tune ll -zerolatency 1 -b:v 6M \
-f flv rtmp://server/live/stream
VOD (quality target)
ffmpeg -i input.mp4 \
-c:v hevc_nvenc -preset p6 -tune hq \
-rc vbr -cq 22 -b:v 0 \
-rc-lookahead 32 -spatial-aq 1 \
output.mp4
Batch / parallel encoding
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input1.mp4 -c:v h264_nvenc output1.mp4 &
ffmpeg -hwaccel cuda -hwaccel_output_format cuda -i input2.mp4 -c:v h264_nvenc output2.mp4 &
wait
Docker (GPU passthrough)
docker run --gpus all --rm -v $(pwd):/data \
jrottenberg/ffmpeg:nvidia \
-hwaccel cuda -hwaccel_output_format cuda \
-i /data/input.mp4 -c:v h264_nvenc /data/output.mp4
Best Practices
- Use full GPU pipelines when possible to avoid CPU/GPU memory transfers
- Match decode and encode hardware for best performance
- Pick presets deliberately - faster is not always better for quality
- Enable lookahead and AQ for quality-critical encodes
- Test on target hardware - quality varies by GPU generation
- Monitor GPU memory for high-resolution content
- Consider power efficiency for laptops and servers
- Update drivers regularly for performance and feature improvements
Reference Map
For deep dives on each backend, see:
- NVIDIA NVENC/NVDEC (presets, CUDA filters, scale_npp, hybrid pipelines, two-pass, multi-GPU, 1:N ABR):
references/nvidia-nvenc.md - Intel QSV (full pipeline, vpp_qsv, lookahead, multi-GPU, VVC decoding):
references/intel-qsv.md - AMD AMF, VAAPI, Apple VideoToolbox (encoders, scaling, VVC VAAPI 8.0+ SCC support):
references/amd-amf-vaapi-videotoolbox.md - Vulkan Video and Vulkan filters (h264/hevc/av1/ffv1, VP9/ProRes RAW decode, scale/overlay/blur/transitions/libplacebo):
references/vulkan.md - OpenCL filters (scale, overlay, colorkey, nlmeans, deshake, unsharp, tonemap):
references/opencl-filters.md - GPU memory management, filter comparison matrix, Docker, troubleshooting:
references/gpu-memory-and-troubleshooting.md
Related Skills
JosiahSiegel/ml-training
development
VerifiedTrustedCommunity
This skill should be used when the user asks to train, debug, scale, or improve ML models. PROACTIVELY activate for: (1) PyTorch, TensorFlow/Keras, JAX, Flax, Hugging Face Trainer/Accelerate training loops, (2) distributed training, DDP/FSDP/DeepSpeed, TPU/GPU setup, (3) mixed precision AMP/bf16, gradient accumulation, checkpointing, seeding, (4) overfitting, imbalance, loss functions, regularization, LR schedules, warmup, (5) memory optimization, gradient checkpointing, offloading, quantization-aware training. Provides: reproducible training best practices across deep learning and classical ML.
38SKILL.mdUpdated May 28, 2026
JosiahSiegel/ml-mlops
development
VerifiedTrustedCommunity
This skill should be used when the user asks to productionize, track, version, govern, monitor, or automate ML systems. PROACTIVELY activate for: (1) MLflow, Weights & Biases, Neptune, Comet, ClearML experiment tracking, (2) model registry, model versioning, artifact lineage, reproducibility, (3) Kubeflow, SageMaker Pipelines, Vertex AI Pipelines, Azure ML pipelines, Databricks workflows, (4) CI/CD, continuous training/evaluation, A/B tests, canary/shadow deployments, (5) drift detection, model monitoring, data validation, responsible AI governance. Provides: end-to-end MLOps architecture and operational safeguards.
38SKILL.mdUpdated May 28, 2026
JosiahSiegel/ml-inference-optimization
development
VerifiedTrustedCommunity
This skill should be used when the user asks to optimize, export, serve, compress, or accelerate ML inference. PROACTIVELY activate for: (1) latency, throughput, p95/p99, batching, concurrency, KV cache, memory, or cost issues, (2) quantization INT8/INT4, GPTQ, AWQ, bitsandbytes, pruning, sparsity, distillation, (3) ONNX export, ONNX Runtime, TensorRT, TorchScript, torch.compile, XLA, OpenVINO, Core ML, TFLite, (4) Triton, TorchServe, TF Serving, BentoML, Seldon, KServe configuration, (5) edge deployment, CPU/GPU/TPU/Inferentia serving. Provides: hardware-aware inference optimization and safe benchmarking.
38SKILL.mdUpdated May 28, 2026
JosiahSiegel/ml-hyperparameter-tuning
testing
VerifiedTrustedCommunity
This skill should be used when the user asks to tune hyperparameters, run sweeps, optimize search spaces, or use AutoML. PROACTIVELY activate for: (1) Optuna, Ray Tune, FLAML, AutoGluon, Hyperopt, Nevergrad, KerasTuner, W&B sweeps, (2) grid search, random search, Bayesian optimization, TPE, Gaussian processes, evolutionary search, (3) ASHA, Hyperband, successive halving, multi-fidelity optimization, population-based training, (4) learning-rate finder, batch-size search, early stopping, pruning, (5) reproducible sweep design and experiment analysis. Provides: budget-aware hyperparameter search strategy.
38SKILL.mdUpdated May 28, 2026