> ## Documentation Index
> Fetch the complete documentation index at: https://docs.openinary.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Video Processing

> Configure the video job queue: concurrency, retries, and cleanup

## Parameters

| Parameter                                                         | Description                            |
| ----------------------------------------------------------------- | -------------------------------------- |
| [`VIDEO_MAX_CONCURRENT`](#video-max-concurrent)                   | Max simultaneous video jobs            |
| [`VIDEO_JOB_RETRY_MAX`](#video-job-retry-max)                     | Max retry attempts per job             |
| [`VIDEO_JOB_CLEANUP_HOURS`](#video-job-cleanup-hours)             | Hours before completed jobs are purged |
| [`VIDEO_WORKER_POLL_INTERVAL_MS`](#video-worker-poll-interval-ms) | Worker polling interval                |

## Parameter reference

<a id="video-max-concurrent" />

<ParamField path="VIDEO_MAX_CONCURRENT" type="integer">
  Maximum number of video jobs processed at the same time.

  When unset, auto-detected from available RAM (1 worker per 2 GB, max 16).

  | RAM   | Auto-detected workers |
  | ----- | --------------------- |
  | 4 GB  | 2                     |
  | 8 GB  | 4                     |
  | 16 GB | 8                     |

  <Warning>
    Each worker spawns an FFmpeg process. Setting this too high will exhaust CPU and memory.
  </Warning>
</ParamField>

<a id="video-job-retry-max" />

<ParamField path="VIDEO_JOB_RETRY_MAX" type="integer" default="3">
  Maximum retry attempts for a failed job before it is marked as permanently failed.

  `3` (default) · `5` (unreliable network) · `1` (fail fast)
</ParamField>

<a id="video-job-cleanup-hours" />

<ParamField path="VIDEO_JOB_CLEANUP_HOURS" type="integer" default="24">
  Hours after which completed and failed jobs are deleted from the database.

  `24` (default) · `168` (1 week, for debugging) · `1` (aggressive)
</ParamField>

<a id="video-worker-poll-interval-ms" />

<ParamField path="VIDEO_WORKER_POLL_INTERVAL_MS" type="integer" default="1000">
  How often the worker checks for new jobs, in milliseconds. Lower = faster pickup, higher CPU. Higher = slower pickup, lower CPU.

  `1000` (default) · `500` (faster) · `2000` (lower resource use)
</ParamField>

## Configuration examples

<Tabs>
  <Tab title="Development">
    ```bash theme={null}
    VIDEO_MAX_CONCURRENT=2
    VIDEO_JOB_RETRY_MAX=3
    VIDEO_JOB_CLEANUP_HOURS=1
    VIDEO_WORKER_POLL_INTERVAL_MS=1000
    ```
  </Tab>

  <Tab title="Production (high performance)">
    ```bash theme={null}
    VIDEO_MAX_CONCURRENT=8
    VIDEO_JOB_RETRY_MAX=5
    VIDEO_JOB_CLEANUP_HOURS=48
    VIDEO_WORKER_POLL_INTERVAL_MS=500
    ```
  </Tab>

  <Tab title="Resource constrained">
    ```bash theme={null}
    VIDEO_MAX_CONCURRENT=1
    VIDEO_JOB_RETRY_MAX=2
    VIDEO_JOB_CLEANUP_HOURS=12
    VIDEO_WORKER_POLL_INTERVAL_MS=2000
    ```
  </Tab>

  <Tab title="Auto-detected concurrency">
    ```bash theme={null}
    # VIDEO_MAX_CONCURRENT omitted — auto-detected from RAM
    VIDEO_JOB_RETRY_MAX=3
    VIDEO_JOB_CLEANUP_HOURS=24
    VIDEO_WORKER_POLL_INTERVAL_MS=1000
    ```
  </Tab>
</Tabs>

## Related

<CardGroup cols={2}>
  <Card title="Video Transformations" icon="video" href="/media-transformations/video-transformations">
    URL parameters for resizing, trimming, and encoding videos.
  </Card>

  <Card title="Upload & Cache Warming" icon="upload" href="/media-transformations/upload-and-prewarm">
    Pre-generate video variants at upload time.
  </Card>
</CardGroup>