← Back to featured case studies

Sky Buddy.
A TV you can talk to.

Millions of people can't really use their television. Voices are too faint, hands don't work the way the remote expects, menus assume you can see and read them easily. I spent a year on a concept where you just talk to it instead.

When
2025, during my year at King's College London
Who
Just me, with five real households
What I did
Research, personas, the voice model, onboarding, prototype, testing
Tools
Figma, tested on an actual TV

This is my own concept project. It isn't affiliated with Sky, and nobody asked me to make it.

Watch first

The prototype, tested on a real television.

The whole thing running, start to finish. The rest of this page is how it got there.

01 / 18

Where it started.

People with speech differences and strong accents feel shut out by voice controls on TV. It mishears them, they don't know where the help settings are, and they end up asking someone else to do it for them. That's the sentence I wrote at the start and kept coming back to.

02 / 18

The problem, and what's inside it.

What's wrong

  • If you have an accent, a cognitive impairment, or you're not confident with tech, voice control fights you.
  • Onboarding is rigid, and it never reassures you.
  • Every time it mishears you, you trust it a little less.

What that opens up

  • Design it inclusively and far more people can use it.
  • People gain the confidence to try voice at all.
  • Being genuinely good at access is worth something to a brand.
03 / 18

I went through the current TV first.

Before designing anything I used the thing as it is, and wrote down every place it broke. Five kept recurring.

  1. Some programmes have no subtitles at all, and where they exist you often can't change the size, colour or position.
  2. It struggles with certain accents, won't wait if you're slow to speak, and talks over you.
  3. There's a setting to slow the voice guidance down. It makes no audible difference.
  4. Customer support is hard to navigate and ignores you if you try to interrupt.
  5. The accessibility remote is more complicated than the normal one, which rather defeats the point.
04 / 18

Then I went to five homes.

For a lot of disabled viewers the television isn't entertainment, it's company and routine and a window out. I met people who needed a carer or a lot of patience just to change the channel. These five shaped everything after.

Tina, 52
Primary progressive MS, bedbound
"I want programmes to play one after the other, so I can watch through the night without help."
Julie, 28
Brain injury, around 60 seizures a week, autistic
"I love that I can go straight to something I've pre-picked and know it'll be something I like."
Jermaine, 43
Low vision, Peters anomaly
"Audio Description shouldn't be a privilege, it's a right."
Arjun, 68
Stroke, hearing loss, one-handed
"It's difficult to watch TV as a family because we all have different needs."
Jake, 7
Learning difficulties, non-verbal, sensory needs
"I can't speak or use the remote, so my Dad helps me every time I want to watch the big TV."
05 / 18

What they said, thought, did and felt.

Said

"Why are there so many options?" · "Show me cartoons!" · "It's not listening!" · "This is very slow."

Thought

I hope I don't press the wrong button. Will it understand my voice? This should be as easy as talking to a friend.

Did

Held the remote carefully, got lost in menus, tried voice quietly, gave up quickly when stuck.

Felt

Frustrated. Anxious about getting it wrong. Relieved when something finally worked.

06 / 18

Forty-one notes on a wall.

Every quote and behaviour from those five visits went onto its own note. I grouped them without deciding the headings first, and let five clusters form on their own. How heavy each cluster was is what set my priorities.

11 notes
It doesn't recognise me

"It never gets my accent." Four of the five had given up on voice search entirely. The system fails, and the person blames themselves.

9 notes
The help is hidden

"I didn't know voice existed." Settings sit four levels down. If you can't find it, it may as well not be there.

8 notes
Afraid of breaking it

"I don't want to break it." People wait for family before trying anything. So when it corrects you, it has to teach, not tell you off.

7 notes
It forgets me

"It forgets me every time." The same setup, every single session. Remembering someone is what earns their trust.

6 notes
One screen, many people

"We leave it as it is, for everyone." A household shares one TV, and everyone in it needs something different.

07 / 18

What I was actually aiming for.

Someone with a physical or speech impairment should be able to find and watch what they want on their own, without asking anyone for help. I mapped one persona across seven questions: what they do, see, say, hear, think and feel. That gave me something concrete to check every decision against.

08 / 18

Turning that into questions I could design against.

  1. How might voice recognition learn how someone speaks, and get better at it over time?
  2. How might setup work out what someone needs and configure itself?
  3. How might there always be another way through when a spoken command isn't understood?
  4. How might it tell you clearly that it heard you, or that it didn't?
  5. How might finding something to watch get simpler for people who find menus hard?
  6. How might someone see and control exactly what it's listening to?
09 / 18

Sixteen ideas, on paper.

Trainer mode. A voice companion. A calm profile. Proximity zones. Group watching. I drew and wrote out every idea before designing anything, so the weak ones could be thrown away cheaply. They're deliberately rough. A rough sketch invites criticism, a polished one invites agreement.

Paper sketches
10–13 / 18

All sixteen, worked up.

Voice-first ideas first, then quieter ambient ones, then ideas about guidance and mood and who else is in the room, and finally the more immersive ones. Click any sketch to open it full size.

Concept 1
Concept 2
Concept 3
Concept 4
Concept 5
Concept 6
Concept 7
Concept 8
Concept 9
Concept 10
Concept 11
Concept 12
Concept 13
Concept 14
Concept 15
Concept 16
14–15 / 18

Mid-fidelity wireframes.

Grey on purpose. At this stage the questions were about order and wording, not colour: does the setup explain itself, is the voice-profile step worth the effort, does anyone understand what "repeat after me" is for.

The home screen, with Sky Buddy offering to set itself up
Choosing a buddy: five voices to pick between
Skyla introduces herself and asks to learn your voice
Repeat a phrase, ten per cent through
Repeat a phrase, forty per cent through
Repeat a phrase, eighty per cent through
Voice setup complete
Your voice has been added to the profile
How often should it keep learning your voice
Playing Sky News, asked for by voice
Back on the home screen, with things you can say

The whole voice setup, end to end: meet your buddy, teach it your voice, then ask it for something.

16–17 / 18

Then the visual pass.

Sky's own type and colour, the buddy given a face, and the whole thing put on an actual television to see whether it still read from across a room.

High fidelity · click through the prototype

High fidelity · the screens

18 / 18

Where it got to.

Design for the edges and everyone gets something. The same voice-first design that frees someone with no use of their hands also helps a parent holding a sleeping baby. Access was the starting point here, not a box ticked at the end.

No remote
Full control by voice, no carer needed
5 voices
From playful to completely calm
Real TVs
Tested in people's actual living rooms
One sentence
To reach subtitles, description or slower speech

The whole thing running, start to finish.