Generating and editing videos – Azure AI applications (AI-103, lesson 18)

What you learn in this lesson

generate videos from text and from a start image with the video model and build the asynchronous flow of job, status query and retrieval in .NET

change an existing video in a targeted way with a remix and say what may change in the process

check limits and content rules of the video model: up front, preview, region, concurrent jobs, retention period

recognize contradictions in the documentation and carry them as open instead of favoring one page

build a queue that keeps the limits and saves every result right away

For a destination image of the catalog (lesson 3.1), the editorial team of Nordlicht Reisen GmbH (R-01) wants a short clip: the ferry enters the harbor, plus a variant in an evening mood (R-05). This lesson shows how a video job runs, what the documentation promises and where it contradicts itself. Paper: requirements, decision, compiled code. Only the optional cloud step needs an Azure account. Cost drivers: video seconds (Learn names billing per second), resolution, number of variants; the pricing page is under "Further reading". All motifs are made up and contain no people (R-12).

New in this lesson. For videos the case fixes a fifth deployment: video (R-08 names four; the name follows the task, not the model). Lesson 2.1 still checks only the first four names.

Prior knowledge. A counter around every model call is in "AI track, lesson T2.4", approvals in "T7.3", hard rules in code in "T6.4". Images and provenance marks are in "lesson 3.1" and "lesson 1.5" of this course. What is new here: a service that needs minutes instead of milliseconds.

The requirement

All numbers are assumptions of the task.

No.	Requirement	Type	Measure	Weight	Source
3.2-01	Short clip from a text, landscape format.	F	1 clip, 8 seconds	Must	Editorial team
3.2-02	Clip from a destination image (lesson 3.1) as the start image.	F	1 start image	Should	Editorial team
3.2-03	Variant with a different color mood, rest unchanged.	F	1 change per remix	Should	Editorial team
3.2-04	Jobs run in a queue; the service's limits are kept.	R	0 rejected jobs	Must	Operations
3.2-05	Every finished clip is saved right away.	R	100 % of clips	Must	Operations
3.2-06	No people, no third-party brands; approval before publication.	R	0 clips without approval	Must	Legal
3.2-07	Video seconds per month are capped.	Q	counter per job (T2.4)	Must	Operations (R-09)

This lesson belongs to course AI-103

You will see the rest of this lesson and its video once you have bought the course. The goals above show what you learn here.

€129.00

incl. VAT, one-time payment

Access to all of the course's lessons with their exercises, solutions and videos on this site, for 24 months from the purchase. A self-study course: no tutoring, no exams or grading by us, and no certificate.

Withdraw from contract here