Generation Time

The total turn-around-time (TAT) for the generation of a didimo is dependant on a number of factors. This page provides guidance as to the likely TAT for different specifications of didimo.

Definition

Didimo defines the TAT for a didimo as

📘
The time between the receipt of a generation request by Didimo servers, and the time that the didimo asset is available for download.

This time does not include the network transfer times both ways.

TAT Variables

The TAT will vary based on a number of variables, namely -

TAT Variable	Description
Input Image (Seflie) resolution	High-resolution input images will generate better results but may take longer to process in the DGP.
The didimo capability specified in the generation request	More complex generations that include more didimo features, or additional processing modules within our pipeline can include slightly higher generation times.
The number of queued didimo requests and accessory fitting requests (Trial and Access Package customer only)	Trial users and Access Package customers have all their generation requests and accessory fitting requests processed by a specific pool of machines. Whilst we can dynamically scale these, at peak times when we may be processing large volumes there may be some TAT delays.
The number of queued didimo requests and accessory fitting requests from the account (Trial and Access Package customer only)	Trial users and Access Package customers may only process two parallel generation requests and accessory fitting requests - i.e., the third concurrent request will be queued.

📘
n.b Enterprise Customers will have dedicated infrastructure and will receive a consistent and predictable TAT.

TAT Predictor

In order to give some predictability as to the approximate TAT that should be expected, as well as the size of didimo package that will be generated, we have provided two reference tables (one for full-body requests and another for head-only requests) below.

Please note that we have not included every single combination of generation variables, but a few select ones to give a likely indication.

For the purpose of this table, all input images were 12.2 MP.

Full-body

	Example 1	Example 2	Example 3	Example 4	Example 5
Input Image (Selfie)	x	x	x	x	x
glTF File Format	x	x	x	x	x
No Poses	x
Simple Poses		x	x	x	x
AWS Polly Poses		x			x
ARKit Poses		x	x		x
Oculus Lipsync Poses			x	x	x
2k Max Texture Dimension (default)		x		x	x
1k Max Texture Dimension
512 Max Texture Dimension	x		x
PACKAGE SIZE	5.3 MB	65.6 MB	10.9 MB	62.6 MB	66.9 MB
~TAT (seconds)	75	80	80	78	81

Head-only

	Example 1	Example 2	Example 3	Example 4	Example 5
Input Image (Selfie)	x	x	x	x	x
glTF File Format	x	x	x	x	x
No Poses	x
Simple Poses		x	x	x	x
AWS Polly Poses		x			x
ARKit Poses		x	x		x
Oculus Lipsync Poses			x	x	x
2k Max Texture Dimension (default)		x		x	x
1k Max Texture Dimension
512 Max Texture Dimension	x		x
PACKAGE SIZE	2.9 MB	36.9 MB	8.4 MB	33.9 MB	38.3 MB
~TAT (seconds)	70	74	72	70	74

Since release 2.5.0 the TAT times went from around 90 sec to 70 sec for head-only didimos.