AWS Elemental Inference function quotas
The following quotas and limits apply to functions. To request an increase to a quota that you can change, see Quotas for AWS Elemental Inference.
The quotas that most often limit a function in practice are not Elemental Inference quotas at all. They are your account's Amazon Bedrock quotas for the model you chose — requests per minute and tokens per minute — because the invocation runs in your account. A function on a short interval is a continuous workload against those quotas. For more information, see Costs and throughput.
Output quotas
The following table lists the quotas for the outputs of a feed.
| Quota or limit | Default | Adjustable |
|---|---|---|
| Enabled function outputs per feed | 2 | Yes |
| Enabled outputs per feed, of all types | 5 | No |
| Total outputs per feed, enabled and disabled | 30 | No |
Every enabled output on a feed counts against the per-feed output quotas, whichever service created it, and a function output counts against both the function-output quota and the all-types quota. A feed that AWS Elemental MediaLive set up for event clipping and smart subtitles is already using two of its five enabled outputs.
Function and expression limits
The following table lists the limits for a function and for each of its runs.
| Quota or limit | Default | Adjustable |
|---|---|---|
| Interval of a fixed-interval trigger | 1–60 seconds | No |
| Frames per run | 1–64 | No |
| Size of a resized frame | Width up to 1,280 pixels, height up to 720 pixels | No |
| Length of each expression | 5,120 characters | No |
| Length of a function description | 1,024 characters | No |
| Length of a model ID or ARN | 2,048 characters | No |
| Size of the request after media substitution | 25 MB | No |
| Model invocation timeout | About 18 seconds | No |
| Time from the end of a window to its result, on a live feed | At most about 20 seconds | No |