SundaySky · Lead UX Designer

Document to video

Turn documents into videos with clarity, control, and trust.

AI feature Editor SaaS Prompt design Q3 2025
A generated welcome video beside the creation menu offering From scratch, Upload a document, With AI Copilot, and From Inspiration Gallery.
4Participants, prototype validation
6Discovery insights
5Success metrics defined

Designed to concept and prototype. Later shipped: see below.

Goal

Enable users to turn documents into videos quickly, while maintaining trust, transparency and creative control throughout an AI-driven flow.

The challenge

Users expect fast, automated results with minimal effort, but still want to understand what the AI is doing, preview outcomes early, and make light refinements before committing to generation. Most tools feel like a black box, which reduces confidence and adoption.

My role
Lead UX Designer: concept, not launched
Team
PMs, Engineers, UX Researcher, Copywriters, CS
Tools
Figma, FullStory, Bolt, Respondent
Timeline
Q3 2025
Process
Discovery → Prototype → Design

Discovery

3 inputs

01. Target user · content / marketing

Wants a branded video out of a document, and wants to see it before committing.

Goal

Create branded, concise, engaging videos from internal documents.

Pain points

Tools often feel like a “black box”: unclear what's happening, while expecting great results from little work.

Tools

  • Canva
  • Synthesia
  • AI Studios
  • Veed
  • ChatGPT
  • Gemini

Needs

Fast preview, light editing, and clarity before committing.

Portrait of the content and marketing persona.

User testing · UX research & Pendo feedback

02. Benchmark & inspiration

Leading tools prioritize speed and transparency over feature depth.

Top document-to-video platforms focus on fast onboarding, early previews and guided AI flows. Tools that delay feedback or hide AI logic reduce trust and adoption, even when their results are visually strong.

  • Canva
  • Synthesia
  • Veed
  • Google AI Studio
  • ChatGPT
  • HeyGen

Competitive benchmark · AI video tools

03. Interviews · requirement exploration

Clear interest in document-to-video, with a strong emphasis on early preview and guidance.

Participants recruited through Respondent, all with prior experience making videos from documents on four known platforms. Each completed the task twice (once with a familiar tool, once with an unfamiliar one) using the same document. That comparison revealed early gaps in trust, clarity and guidance, and shaped the initial product insights.

Comparative interview findings across familiar and unfamiliar document-to-video tools.

Interviews & observation · UX researcher · 4 participants

What discovery told us

  • Trust: users trust the process when they can see early results.
  • Preview: the preview-first flow is the key differentiator.
  • Copilot guidance: the Copilot should act as a guided assistant, not a chatbot.
  • AI avatar: avatars enhance engagement but should stay optional.
  • Iteration: users want light iterations before committing to a full generation.
  • Terminology: clear terminology is essential for confidence.

Prototype

Validated with 4 users

Validation · flow testing

The preview-first flow increased confidence and reduced hesitation throughout the process.

A prototype was tested with four users: two internal, two external.

What worked

  • Preview built trust and confidence.
  • The upload flow felt intuitive and clear.

What didn't

  • The Copilot felt static; users wanted richer guidance.
  • Terminology (“Audience”, “As-Is”) caused confusion.
Prototype validation findings on trust, upload clarity, Copilot guidance and terminology.

Prototype validation · 4 participants

Design

Flow and 5 solutions

Final UX flow

Upload, set, preview, edit, and only then generate.

Final flow: upload a document, choose settings, see an interactive preview, make light edits, then generate.

Design system

Worked inside SundaySky's system, adding only what the AI flow needed.

  • Worked within SundaySky's design system.
  • Added new components for the avatar placeholder.
  • Added a split-screen AI Copilot assistant view and its interaction.
Design system sheet: components, buttons, notifications, icons, typography and colours.

Solution

Solution 01. Upload file

Decision: analyse the structure, then offer a choice

Users upload documents and decks; the system auto-analyses structure (slides, sections, hierarchy), and offers “As-Is” and “Generate New” modes. A clear entry point, with the file replaceable or removable at any time.

The Document to video dialog: drop a single document or browse, with a 40MB limit stated.

Solution 02. Video settings

Decision: give control before generation, not after

Early control over tone, voice, brand and length, with defaults recommended from the detected content and credit usage shown up front. Deciding before generating is what makes the output feel owned rather than delivered.

Video settings: goal, length, tone options and target audience, with a scene preview alongside.

Solution 03. Video preview

Decision: preview first: this is the whole idea

An interactive visual preview before the final output, with quick edits to narration, tone and scene, and simple options for small changes. Instant, editable feedback is what turns an AI pipeline from a black box into something a user will trust.

The interactive preview: script and scene list on the left, playable video preview on the right.

Solution 04. Banner entry point

Decision: make the AI flow more inviting Scoped out of MVP

An alternative entry point designed to make the AI-powered flow more inviting and guide users to explore it. Cut from MVP scope.

A banner asking how you would like to create a video, offering from scratch, from a document, from a script or from a URL.

Solution 05. AI Copilot upload

Decision: structured prompts to guide the whole flow Out of scope

Structured chat prompts to guide the flow, with a “generate” shortcut, step-by-step refinements and context kept visible while editing. Out of scope for this phase.

The AI Copilot composing a video brief through structured prompts, with settings along the bottom.

Outcome

Concept phase

What we would have measured

Five metrics defined before launch, so the concept could be judged rather than admired.

Success metrics

  • Preview interaction.
  • Time to first preview.
  • Iterations before generation.
  • Flow completion rate.
  • Drop-off during setup.

Key trade-offs

  • Speed, clarity and early feedback were prioritized over advanced editing and deep customization.
  • That reduced cognitive load and increased confidence, but limited flexibility for power users in the early stages.

Project insights

It ended at concept and prototype, so there are no numbers, and that's worth saying plainly.

This project ended at the concept and prototype phase, before launch, so quantitative analytics were not available. Insights from research and validation showed that early preview is the strongest trust builder, while clearer terminology and richer Copilot guidance are critical for confidence.

Next step

Ship an MVP, iterate on guidance and language, and validate whether preview-first flows actually drive adoption and completion.

Impact · it shipped

The concept didn't launch on my watch. It launched.

I designed this to concept and prototype and left SundaySky in 2025. The feature is now live and marketed as Doc-to-video, part of SundaySky's prompt-to-video offering: carrying forward the flow validated here: upload a document, get an editable draft, refine before committing.

“Quickly and easily convert existing word documents, .txt files, and PDFs into editable draft videos that teams can refine and share in minutes.” SundaySky, Business Video Creation · product page, 2026
Doc-to-video in SundaySky: a document uploaded and converted into an editable draft video. Play · product footage, 1.8 MB

Product footage and copy: SundaySky, 2026 · sundaysky.com. The shipped interface postdates my involvement and may differ from the concept above.