TL;DR
Open the Klar3D Point Cloud Converter, select multiple files of the same supported input format, choose one output format, and start the batch. Each original file is converted and saved before the next file is read.
Conversion runs locally in your browser. No account, upload, or installation is required.
1. Prepare files with the same format
Group your PCD files together, your LAZ files together, or another supported static format. The general converter reads and writes PCD, PLY, LAS, LAZ, XYZ, CSV, and PTS, with available outputs determined by the input.
Each batch uses one input format and one output format. For example, several PCD files can become separate LAZ files. Conversion does not align or merge those files into a single spatial dataset. MCAP recordings use their own frame-selection workflow rather than this static-file batch queue.
2. Select the batch
Select or drop several same-format files. Review the queue and the first file’s metadata and preview. The first preview is a useful check of the input choice, but it does not establish that every other source is valid or has identical attributes.
For a first run, try two small representative files before selecting a large collection.
3. Choose the output and review warnings
Select an available output format. Choosing XYZ, for example, cannot retain RGB or intensity because this XYZ writer exports coordinates only. LAS/LAZ output does not preserve every source metadata field, and ordinary LAS/LAZ input is limited to LAS 1.0-1.3 point formats 0-3.
Keep source files and any coordinate-system documentation. Conversion does not infer units or reproject a dataset. For one specific pairing, see LAZ to PCD; for text inputs, see XYZ, CSV, and PTS to LAZ.
4. Choose where results are saved
Where the browser supports it, use Save batch to folder and select a destination. The converter creates a timestamped run folder for the results.
Other browsers use ordinary downloads. You may need to allow multiple downloads when prompted. This workflow produces separate outputs; it should not be confused with the ZIP packaging used for multi-frame recording exports.
5. Run and review the queue
Start conversion and watch the queue’s Waiting, Converting, Saved, or Failed states. Processing is serial: the current output is saved and released before the next source is read. That limits how much batch data is held at once, although a single large file still needs sufficient browser memory.
Review the final saved and failed counts. Check failed inputs individually, and reopen representative outputs to compare point counts, bounds, and expected fields. A completed batch should be assessed by its saved results, not only by how the first preview looked.
Working with recordings or scans
To extract a selection of recorded LiDAR frames, follow MCAP to LAZ or PCD or Ouster PCAP export. To select scans inside an E57 container, use the E57 prototype tutorial.