Experiment: Google Veo 3 and Branching Scenarios

What question did we test?

We tested whether Google Veo 3 could replace traditional video production for a branching scenario, giving learners a believable person who reacts to their choices. The specific challenge was consistency: a branching scenario needs the same character across many clips, and AI video tools often change a face from one generation to the next.

Branching scenarios are one of the best formats for soft skills, because learners see consequences instead of reading about them. Their weakness has always been production. Filming every branch with actors is expensive, and stock footage rarely fits. When Veo 3 arrived, we wanted to know if it closed that gap.

How was the experiment set up?

We chose a situation every client-facing team knows: a client asks for an extra round of feedback beyond what the contract includes. We generated the client’s video in Google Veo 3, framed it to look like a real Zoom call, and built the decision points in Articulate Storyline. Tested in July 2025.

What did we build?

We built an interactive branching video module where the learner, as an account manager, responds to Julia’s request. At the decision point, learners choose one of three ways to respond: agreeing too quickly, pushing back abruptly, or setting professional boundaries empathetically. Each choice plays Julia’s reaction.

1:26

The key technique was character consistency. We wrote one visual prompt for Julia and reused it for every take, changing only the audio dialogue. That kept her natural expressions, gestures, and tone stable across the clips, which is what makes a branching scenario believable. The walkthrough at the top of the page shows the full breakdown, including the Storyline build.

What happened when we tested it?

The approach worked. Reusing a fixed visual prompt kept Julia consistent enough that the scenario felt like one conversation, and the realistic Zoom framing made the stakes feel familiar. The main limits were the ones every AI video build still has: careful prompting, review of each clip, and planning all branches before generating anything.

What worked

• Fixed visual prompts with changing dialogue produced a consistent character.
• Realistic reactions made weak choices visibly costly, which is the point of the exercise.
• Storyline handled the branching logic with standard features.
• No actors, studio, or reshoots were needed to produce the variations.

What needed work

• Each clip needed review for small inconsistencies [VERIFY: number of takes discarded].
• Branches had to be fully scripted before generation, because regenerating later risks drift.
• Veo clips are short, so longer reactions need planning (we later handled this with idle loops in our customer service experiment).

This walkthrough became one of the most watched experiments on our channel, with over 2,000 views, which told us many learning teams are asking the same question.

Should you use Google Veo for branching scenarios? The verdict

Google Veo is a strong fit for branching scenarios when a realistic reaction teaches the lesson and the paths can be scripted in advance. It is a weaker fit when scenes need precise physical actions, many characters interacting, or long continuous takes, and when your organization requires real employees or actors on screen.

If you want unscripted practice instead, compare our AI roleplay training simulator.

How can you try a Google Veo branching scenario yourself?

Start with one decision point and three choices, one strong and two common mistakes. Script every line first, write one fixed visual prompt for your character, generate each reaction with only the dialogue changing, and build the branching in your authoring tool. Test with people who face the situation for real.

1. Pick a familiar, uncomfortable moment. Scope creep, a complaint, or a pushy request.
2. Script three responses. One strong choice and two realistic mistakes.
3. Write the character prompt once. Appearance, setting, lighting, framing.
4. Generate reactions by changing only dialogue.
5. Review every clip for drift in face, voice, and setting.
6. Build the decision points in Articulate Storyline or a similar tool.
7. Test with practitioners and refine feedback for each path.

See more finished work in our interactive video examples and Articulate Storyline examples.

What’s next for this experiment?

We already took the next step: our realistic AI video for customer service training experiment applied the same approach to an emotional de-escalation scenario and added Director’s Notes and idle loops. The open question now is how far branching can scale with AI video before consistency breaks [VERIFY: planned multi-scene test].

We’ll publish that in the Lab, along with how these builds compare to fully animated approaches tested in our AI soft skills video production experiment.