--- title: "Create a Job" description: "Submit a transcoding job: the source URL or S3 object, the destination bucket and credentials, the outputs, and the webhook and idempotency key worth setting." category: "Deployment" language: en source: https://cubepath.com/docs/products/deployment/transcoding-jobs updated: 2026-11-17 --- # Create a Job Open **Video Transcoding** → **New Job** in [my.cubepath.com](https://my.cubepath.com). A job is one source, one destination and one list of outputs. There's no project, profile or preset to configure first, and nothing carries over between jobs: every submission brings its own source, its own destination and its own credentials. That's deliberate: a job is fully described by what you send, so you can generate jobs from code without any stored state to keep in sync. ## Source Choose the source type first, because it changes the rest of the form. - **Download URL**: an `http://` or `https://` URL we can fetch without credentials. The form rejects anything that isn't a full URL, and the service refuses URLs that resolve to private or internal addresses, so a link to something on your own LAN won't work. - **S3 bucket**: an object in a bucket you control. You supply the endpoint, region, bucket and **Object key**, plus an access key and secret key. The object key is the path inside the bucket (`videos/talk-42.mov`), not a URL. Whatever the source, it's read once. Ten outputs don't mean ten downloads. ## Destination The destination is always an S3-compatible bucket, and it's the only place the job writes. Fill it in even if it's the same bucket the source came from: input and output are configured separately. | Field | Required | Notes | |-------|----------|-------| | **Endpoint** | No | Leave empty for AWS S3. Set it for CubePath Object Storage, R2, B2, Wasabi or MinIO | | **Region** | No | `auto` is fine for providers that ignore it | | **Bucket** | **Yes** | The one field the form always insists on | | **Destination prefix** | No | e.g. `transcoded/`. Everything the job writes lands under it | | **Access key** / **Secret key** | In practice yes | Only omit them if the bucket accepts anonymous writes; almost none do | > Important: Give the job a key that can read and write under the destination prefix and read the source object, and nothing else. The secret is stored encrypted and never returned by the API, but a scoped key is still the difference between a leaked job config and a leaked account. ## Outputs Add at least one output; a job with none is rejected. Each output is an independent rendition of the same source, and you can mix types freely: an MP4 for downloads, an HLS ladder for playback, thumbnails for a scrubber, a GIF for the card preview. The types and their parameters are covered in [Outputs](/docs/products/deployment/transcoding-outputs). ## Advanced Two optional fields, both worth setting: - **Webhook URL**: we POST to it when the job completes or fails, so your pipeline doesn't have to poll. Delivery is best-effort and isn't retried; see [Job status](/docs/products/deployment/transcoding-job-status). - **Idempotency key**: a unique string of your choosing. If the same key is submitted again, you get the existing job back instead of a second one. Set it any time the caller might retry: a queue worker, a webhook handler, a user double-clicking **Submit job**. ## Submit ```steps ### Pick the source Choose **Download URL** or **S3 bucket**, then fill in the URL or the bucket and object key. ### Fill in the destination Endpoint, region, bucket, prefix and the credentials that can read and write there. ### Add your outputs One row per rendition. Start with a single MP4 if you're testing the path end to end. ![Transcoding output with type, container, codec and height](https://cdn.cubepath.com/img/docs/products/52b960d0-ddf6-4853-b3cc-2a55f9ae77e6.png) ### Submit Click **Submit job**. The job appears in the list as **Queued** and starts as capacity frees up. ``` > Tip: Run one job with a single small file and one MP4 output before you wire anything up. It proves the credentials, the endpoint and the prefix in about a minute, and every later problem is then a real problem rather than a typo in a bucket name. ## When the form pushes back | Message | What to fix | |---------|-------------| | Enter a valid download URL | The URL source needs a complete `http(s)://` URL | | Enter the input object key | An S3 source needs the object key as well as the bucket | | Destination bucket is required | The destination bucket is empty | | Add at least one output | Every job needs one or more outputs | | "Cannot target a private/internal address" | A URL or S3 endpoint points at a private IP; use a public hostname | | The form is valid but the job is refused on submit | The organization is unverified, suspended, or out of balance |