> For clean Markdown of any page, append .md to the page URL. > For a complete documentation index, see https://docs.nvidia.com/nemo/curator/llms.txt. > For full documentation content, see https://docs.nvidia.com/nemo/curator/llms-full.txt. > For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/nemo/curator/_mcp/server. > Transcode clips with CPU or GPU encoders and tune for quality and performance # Clip Encoding Convert extracted clip buffers into compressed media files suitable for storage and training workflows using encoders. NeMo Curator provides both CPU and GPU-based encoders. ## Use Cases * Convert raw clip buffers into a standard format (such as H.264 in MP4) for portability. * Normalize heterogeneous inputs (encoding formats, bit rates, containers) into a consistent output. * Reduce storage footprint with controlled quality settings. ## Before You Start If you only need embeddings or analysis and do not require saved media files, you can skip encoding. When writing clips, NeMo Curator produces `.mp4` by default. *** ## Quickstart Use the pipeline stage or the example script flags to encode clips with CPU or GPU encoders. ```python from nemo_curator.pipeline import Pipeline from nemo_curator.stages.video.clipping.clip_extraction_stages import FixedStrideExtractorStage, ClipTranscodingStage pipe = Pipeline(name="transcode_example") pipe.add_stage(FixedStrideExtractorStage(clip_len_s=10.0, clip_stride_s=10.0)) pipe.add_stage(ClipTranscodingStage(encoder="h264_nvenc", encode_batch_size=16, encoder_threads=1, verbose=True)) pipe.run() ``` ```bash python -m ray_curator.examples.video.video_split_clip_example \ ... --transcode-encoder h264_nvenc \ --transcode-use-hwaccel \ ``` ## Encoder Options | Encoder | Hardware | Description | | ------------- | ------------------ | -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | `h264_nvenc` | NVIDIA GPU (NVENC) | Uses NVENC for high-throughput H.264 encoding on NVIDIA GPU hardware. | | `libvpx-vp9` | CPU | VP9 software encoder. Use as a fallback on GPUs without NVENC silicon (such as A100/H100). Slower than NVENC; produces VP9 in `.mp4`. Emits a one-time performance advisory at construction. | | `libopenh264` | CPU | H.264 software encoder. **Not bundled with Curator's default FFmpeg build** — opt-in by running `bash /opt/Curator/docker/common/install_h264_support.sh --with-libopenh264` inside the container, or providing a system FFmpeg that includes it. The stage probes at setup time and raises with a docs link if missing. See [Software H.264/HEVC/AV1 Codec Support](/get-started/installation#software-h264hevcav1-codec-support-advanced). | On systems with supported NVIDIA GPU hardware and an `ffmpeg` build with NVENC, `h264_nvenc` can significantly increase throughput. Refer to the verification steps below to confirm NVENC availability. On GPUs without an NVENC encoder block (such as A100 and H100), use `libvpx-vp9` instead — it runs entirely on CPU and has no proprietary licensing constraints. **Need software H.264 (libopenh264 / libx264)?** Curator's default FFmpeg build excludes them for licensing reasons. The fastest path inside the container is the bundled installer: ```bash bash /opt/Curator/docker/common/install_h264_support.sh --with-libopenh264 ``` For other paths (system FFmpeg, custom image rebuild) see [Software H.264/HEVC/AV1 Codec Support](/get-started/installation#software-h264hevcav1-codec-support-advanced). **Got `SoftwareCodecMissingError` on an h264 input?** Curator's CPU-only `VideoReader` and `ClipWriter` stages call `ffprobe`, which needs a software `h264` decoder — not available in the default NVDEC-only build. Install software decoder support with `bash /opt/Curator/docker/common/install_h264_support.sh`, or pick a transcode encoder whose output codec the system FFmpeg can software-decode (e.g., `--transcode-encoder libvpx-vp9`). See [Software H.264/HEVC/AV1 Codec Support](/get-started/installation#software-h264hevcav1-codec-support-advanced). ### Verify `ffmpeg`/NVENC Support To use `h264_nvenc`, confirm that your `ffmpeg` build includes NVENC support and install the GPU drivers: ```bash ffmpeg -hide_banner -encoders | grep nvenc ffmpeg -hide_banner -hwaccels | grep -i nv nvidia-smi ``` Expected output includes entries like `V..... h264_nvenc` and `cuda` in the hardware accelerators list. If not present, install an `ffmpeg` build with NVENC and ensure NVIDIA drivers and CUDA are available. ## Configure Use `ClipTranscodingStage` to control encoder choice, batching, and acceleration: ```python from nemo_curator.stages.video.clipping.clip_extraction_stages import ClipTranscodingStage transcode = ClipTranscodingStage( encoder="h264_nvenc", use_hwaccel=True, # enable NVENC when using h264_nvenc encoder_threads=1, # CPU thread count for CPU encoders encode_batch_size=16, # number of clips per encode batch num_clips_per_chunk=32, # chunking for downstream writing use_input_bit_rate=False, # set True to preserve input bit rate when available num_cpus_per_worker=6.0, verbose=True, ) ``` ### Parameters | Parameter | Description | | -------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | `encoder` | Selects the encoding backend. Supported values: `h264_nvenc` (GPU, requires NVENC), `libvpx-vp9` (CPU fallback for non-NVENC GPUs such as A100/H100), and `libopenh264` (CPU; opt-in via `install_h264_support.sh --with-libopenh264` — see [Software H.264/HEVC/AV1 Codec Support](/get-started/installation#software-h264hevcav1-codec-support-advanced)). | | `use_hwaccel` | Enable when using `h264_nvenc`. Not valid with `libvpx-vp9` or `libopenh264`. | | `encoder_threads` | CPU threads per worker for CPU encoders. Increase to use more CPU. | | `encode_batch_size` | Batching size for clips; larger batches can improve throughput. | | `use_input_bit_rate` | If True, attempts to reuse the input bit rate; otherwise, the encoder uses its default rate control. | Refer to the quickstart options in [Get Started with Video Curation](/get-started/video) for command-line flags `--transcode-encoder` and `--transcode-use-hwaccel`. ## Troubleshooting * "Encoder not found": Your `ffmpeg` build may lack the encoder; verify with `ffmpeg -encoders`. * "No NVENC capable devices found": Install NVIDIA drivers/CUDA and ensure the GPU is visible in `nvidia-smi`. * Output mismatch or low quality: Revisit encoder defaults; set explicit bit rate/quality settings as needed, or enable `use_input_bit_rate`.