Computer vision answers · Video Analytics

How do I process many long videos at once with computer vision?

Without running them one by one on your own machine.

Short answer

Build a Workflow with your model in Roboflow for free, then send every video to Roboflow Batch Processing, which runs the Workflow on all of them in the cloud and returns the results as JSONL or CSV. Set a maximum frame rate so long videos only process the frames you need. Most monitoring questions do not need every frame, so a low maximum frame rate, such as 5 per second, cuts the time and cost of long videos without missing what matters.

Start free
Roboflow agent · free to start

Skip the setup. Describe your images.

Describe your images and what you need to find. The agent sets up labeling, training and deployment with you.

Key facts

What you need
Your video files and a Roboflow Workflow with your model
What you get
Results for every video, exported as JSONL or CSV
Video sources
Uploaded files, or S3, Google Cloud Storage and Azure Blob
Runs on
Roboflow cloud machines that scale for each job
Dashboard displaying six video feeds from diverse locations including a warehouse, factory, street, and store, with a status panel indicating 4 videos processed and 2 running.

How to process many long videos, step by step

Five steps from a folder of videos to results for every frame you need.

  1. Connect Roboflow to your coding agent, ChatGPT or Codex

    Add the Roboflow MCP server to Claude, Cursor or Gemini CLI, or install the Roboflow plugin in ChatGPT or Codex. Your assistant can then build the Workflow, start the batch job and fetch the results from a prompt, at no extra cost beyond your Roboflow plan.

  2. Choose or train a model for what you are looking for

    Use a pretrained model if it already finds what you need. Otherwise, take frames from your own videos, label them, and train an RF-DETR model with Roboflow Train.

  3. Build a Workflow and test it on one short clip

    In Roboflow Workflows, add your model and the steps you need, such as tracking or counting. Run it on one short clip and check the results before you process every video.

  4. Run every video through Batch Processing

    Stage your folder of videos, then start one Batch Processing job that runs the Workflow on all of them, with a maximum frame rate. Videos stored in S3, Google Cloud Storage or Azure Blob Storage can be read from there directly.

    Shell
    pip install inference-cli
export ROBOFLOW_API_KEY="YOUR_API_KEY"

inference rf-cloud data-staging create-batch-of-videos \
  --videos-dir ./videos --batch-id site-videos

inference rf-cloud batch-processing process-videos-with-workflow \
  --workflow-id your-workflow --batch-id site-videos \
  --machine-type gpu --max-video-fps 5

    Rather not write code? Ask the Roboflow agent to set this up for you

  5. Download the results

    Check the job's status, then export its output as JSONL or CSV. You get results for each video, ready to load into a spreadsheet or database.

Why 2 million engineers build on Roboflow

  • Label faster

    Auto-label with SAM 3 or your own model, then fix the misses instead of drawing every box.

  • RF-DETR

    Roboflow’s open-source real-time detection model, built for accuracy (opens in a new tab) on small objects.

  • Logic built in

    Counting, zones, tracking and alerts are ready-made Workflow blocks, so there is less code to write.

  • Your model, your weights

    Use your trained weights on every plan, including free, through the open-source Inference package.

  • Run it anywhere

    Self-host on a workstation, Docker, an on-prem GPU or an edge device, or call the cloud API.