Picsart APIs run in the cloud that is elastic and can quickly be scaled up for extra load, and are engineered to serve a global audience through a network of worldwide servers designed to minimize latency.
The typical response time for a Picsart API can be broken down into the following parts
- Whether the API is computationally intensive (Upscale, Ultra Upscale) or computationally simpler (Remove Background, etc)
- Input File size and Format. As the file size increases, there is a corresponding increase in processing time. Similarly, some formats (PNG) are larger than other formats (JPEG) and therefore take a longer time to upload or process.
- Final output resolution (e.g. 1024 px vs 768 px) & the final output type (e.g. PNG, JPEG etc) have a bearing on the response time as well. For example, 16K output images may run into hundreds of megabytes, and delivering such images will take longer time.
- Many of the APIs are powered by AI technology which may have relatively unpredictable processing times including different processing times for different images. The sheer number of parameters for each service and the expected/allowed values for those parameters are also large enough that it will affect the processing time.
Given so many variables, it is hard to characterize response time as a single number or come up with an SLA that can be applied to APIs as a whole.
However, when requested, Picsart can provide benchmark data for each service for preset scenarios that can provide a relative understanding of the performance of each service.
You can refer to this KB Article for further performance improvements to response times: Best practices to further improve API performance