Hollr Case Study

Scalable AI-Powered Video Personalisation

AI is everywhere: in headlines, conferences, and product roadmaps.

Yet for many businesses, it's still an abstract concept, powerful but difficult to implement.

We created Hollr to change that. It enables organisations to generate personalised video messages using generative AI, turning complex technology into a simple, scalable solution for smarter customer engagement.

At Sputnik, we're always looking ahead to future technologies and how they can benefit both us and our clients. Hollr makes AI-driven personalisation simple, helping businesses integrate tailored video content into their workflows without the usual technical hurdles.

Background

Audiences today expect content that is clear, concise, and relevant. With attention spans shrinking — particularly among younger demographics — the effectiveness of long-form written communication is diminishing.

Hollr addresses this shift by enabling organisations to deliver personalised video messages that are easy to understand, engaging and accessible.

Whether it's an insurance quote, an appointment reminder or a subscription renewal, Hollr helps businesses communicate with clarity and impact.

Potential use cases for personalised video:

  • Insurance Quotes

  • Loans & Mortgages

  • Investment Updates

  • Chasing Late Payments

  • Appointment Reminders

  • Welcome Messages

  • Flood Warnings

  • Product Demos

The Journey

We approached Hollr's development by mapping the complete user and system flow. This included external service integrations, approval workflows, and message delivery triggers. By visualising dependencies and potential bottlenecks, we optimised both performance and maintainability.

User actions

Request a video

Form populated for video generation

Middleware

A key element in building Hollr was creating a single piece of middleware that connects external services, decision points, and system processes.

This enables Hollr to remain agnostic with regards to the text-to-video integrations, future proofing it as the LLM landscape continues to evolve.

The middleware allows for fast, reliable mobilisation and integration with a wide range of systems, de-risking the implementation of personalised video messaging at scale.

Result

Video generated result page

Middleware

Middleware

A key element in building Hollr was creating a single piece of middleware that connects external services, decision points, and system processes.

This enables Hollr to remain agnostic with regards to the text-to-video integrations, future proofing it as the LLM landscape continues to evolve.

The middleware allows for fast, reliable mobilisation and integration with a wide range of systems, de-risking the implementation of personalised video messaging at scale.

  • Endpoint Validation

    Sample Video Endpoint Check

  • Code Validation

    Check Codes Sheet for validity

  • Insert Request

    Insert into Dynamo DB Videos Table

  • Request Video Creation

    Create Video

  • Insert message into check queue

    Insert into SQS Check Queue with initial timeout value

  • Creating Video Result

    Return Video Creation Result

Result

Video generated result page

Engineered for success

  • Checkmark

    AI agnostic platform providing flexibility in a changing LLM landscape.

  • Checkmark

    End-to-end service, including design, data ingestion, video generation and delivery.

  • Checkmark

    Demonstrable Proof of Concept and MVP removes uncertainty around feasibility and cost of delivery.

  • Checkmark

    Trackable KPIs to measure business impact.

Project Takeaways

  • Prices

    The AI avatar mispronounced prices, especially those including 'pence' or very high figures. We initially tried spelling out the numbers in a more human way, like 'forty-three pounds' instead of £43, but this was hit or miss. Ultimately, rounding all numbers to two decimal places (e.g., £43.00) resolved the issue.

  • Dates & Names

    We humanised dates so instead of 'Your renewal is due on the 25/02/01', Hollr says 'the first of February'. We also ensured that trickier names like 'Strauß', 'Göring' and 'Siobhán' were pronounced correctly.

  • Sentiment

    The AI video's sentiment often appeared 'angry' or had strange facial expressions for certain phrases, so we adjusted the script to avoid emotive words that triggered stronger speech and odd facial expressions.

  • Product Demo

    As a new service, it's understandable that businesses are skeptical of the feasibility, quality and accuracy of personalised video. To overcome this, we built a 'Try it free' feature, allowing users to generate their own video to experience Hollr first hand.

  • Dynamic Scripting

    It's not as simple as 'mail merging' variables into a script. In order to reflect a natural language flow, the script itself needs a number of variations depending on the data for each case.

Discover hollr

Experience AI-powered personalised video messaging, visit hollr.io and try it end-to-end.

Try Hollr for free
Hollr's Personalised Video Ready emailHollr's Personalised Video Ready email
Case Study

Axelos

Powering best practice

Case Study

Swinton Insurance

Enhancing digital performance

Case Study

Acino

Providing a platform for growth

Let's work together.

Call our Managing Director Andy on 0800 680 9777 or fill in our to discuss how we can help.