OpusClip case study
Turning a creator's old footage into a brand-new story
A 0→1 AI feature for OpusClip, the #1 AI video clipping tool with 16M+ users. The feature helps creators recombine existing clips into fresh, AI-generated videos.
Role
Product Designer
Team
4 designers, 2 leads
Duration
2 months
Year
2024–2025
00
At a glance
The brief asked for AI video generation in OpusClip's voice. Research with five creators surfaced a quieter need. They didn't want to generate from nothing. They had 100+ clips in their library and wanted help reusing them.
We pivoted to a Footage-to-Video flow: AI search, AI storyboard, voiceover editing. Each step balanced AI automation with user control.
16M+
OpusClipusers
#1
AI clippingtool
10
creatorstested
0 → 1
newfeature
01
The Context
In 2024, every video tool was racing toward AI generation. Sora explored full video generation. Caption and CapCut leaned into smart editing. But across the board, AI-generated video was fighting two old enemies: consistency and control.
OpusClip, the #1 AI video clipping tool with 16M+ users, asked us to design what their version of this future could look like.
02
The Initial Direction
We started where the client wanted us to: Script → Video. Users would type a prompt, and AI would generate a script, then a storyboard, then a rough cut. A familiar shape, but one we wanted to pressure-test before going deeper.
Hypothesis
Creators will pay for speed. A prompt-first flow that produces a publish-ready video in one go.
03
Putting it in front of real creators
We ran concept-testing sessions with content creators, vloggers, a digital marketer, and a podcaster. Five creators, five different workflows, one consistent reaction.
The prompt-first flow had some sparkle, but it kept hitting the same wall.
What we got right
AI for efficiency: creators want speed. The storyboard preview was great for quick scanning and rough edits.
Where Script-to-Video fell short
Unreliable and prompt-heavy. Limited control over scene composition and storyline. Outputs not polished enough to publish.
The hidden opportunity
Creators already had 100+ clips. Their content followed a defined personal style. They wanted to reuse old footage to make fresh content.
04
The Pivot
We reframed the question.
How might we
Use AI to transform creators' existing footage into compelling new videos with fresh storylines?
The reframe pulled the work toward two key features.
Fast keyword-based search
Surface the most relevant clips from a user's library in seconds, driven by natural-language prompts.
AI storyline generation
Let AI weave the selected clips into a new story, with a generated voiceover the creator can refine.
05
Design Explorations
We needed to learn fast. We split the whole feature into two opposing design directions: automation versus customization. One leaned into AI taking the wheel. The other into giving creators control at every step.
Automation
The AI does the heavy lifting. Users describe what they want, and the product surfaces, generates, and edits.
AI Recommendation
- AI recommends video clips based on recent videos through algorithms.
Automated generate & edit
- One-tap generating story.
- Drag and hover to edit.
- Generation and editing fused at one step.
Customization
Users stay in control. Every step is editable, arrangeable, and manual.
Add Video to Cart
- Easily add or remove videos as needed.
- Provides a more intuitive way to manage selected footage.
Craft Zone
- Manually select and arrange the transcripts or scenes.
- Users can select only the desired parts to form the final version.
- Fine-tune editing video.
06
Usability Testing
We ran usability sessions with five creators, comparing each surface's two directions: automation versus customization. Four findings came out of the testing. One proposal won each comparison.
Refined question
How can we balance AI-powered convenience with user control across searching, compiling, and editing footage into videos?
Users preferred a relevant clips recommendation.
Proposal 01
Relevant clips surfaced on first load — autosaved projects and AI-suggested topics.
Proposal 02
A more intuitive page with more length and ratio controls.
Why this proposal won
- Reduced cold-start time. Users dove in without typing prompts.
- Mirrored how creators think about their library: recent work and topics close to past content.
A horizontal cart kept the interaction consistent.
Proposal 01
Cart on the side, aligned with the storyboard.
Proposal 02
Cart docked at the bottom. Intuitive, no overlap with results, drag or select.
Why this proposal won
- Side cart fought with the suggested-clips area for space. Bottom cart stayed out of the way.
- Drag-or-select felt closer to the timeline-editing model users already knew.
A simple one-step page beat a multi-step generator.
Proposal 01
Generation and editing fused into one step.
Proposal 02
Generate first, then refine in a separate step.
Why this proposal won
- Switching pages broke the creative flow. Users wanted to keep momentum.
- Inline editing made the storyboard feel like a starting point, not a final output.
Automation won over manual drag-add.
Proposal 01
Regenerate by prompt or delete a section.
Proposal 02
Manual drag-add from a left panel.
Why this proposal won
- Re-running a prompt felt faster than hunting for the right phrase to drag in.
- Deleting unwanted sections matched how users already trim text: direct and obvious.
07
The Refined Flow
Same shape as before, with a user-control checkpoint at every AI step. The refined flow keeps AI doing the heavy lifting and bakes in three deliberate handoffs: select clips, regenerate storyline, edit voiceover.
Refined UI
How do we help creators make those adjustments?
The flow above marks three checkpoints where AI hands the wheel back to the creator. Each one needed UI that helped users adjust the output confidently. We introduced three enhancements, one for each checkpoint.
01
Regenerate storyboard
Re-roll a single scene or the whole storyline without losing existing edits.
02
Add-Clips pop-up
Bring in extra footage mid-flow without breaking the current storyboard.
03
Trim bar
Manual fine-tuning of any clip for the moments AI gets close but not quite right.
08
The Final Solution
AI does the work. The creator stays in the chair. Here is the end-to-end prototype, prompt to final cut, in a single flow.
09
What I learned
The biggest design choice wasn't which UI to ship. It was which problem to solve.
01
Listen past the client's prompt
The brief said AI generation. The research said help me reuse what I already have. Sitting with that mismatch is where the real product was hiding.
02
Convenience needs a control valve
AI users don't want full autopilot. They want to feel like they're driving. Designing the handoff between AI and user was the hardest, most rewarding part of the project.
03
Parallel A/B exploration buys speed
Splitting each surface into two opposing directions let us argue with screens instead of opinions. Usability sessions then picked the winner: fast, defensible, low ego.
04
Constraints are gifts
A two-month timeline and a giant existing user base forced us to design within OpusClip's vocabulary, not invent a parallel one. The feature felt native because it was.
10
Next steps
A few directions worth picking up next.
Storyboard history
Let users revisit past versions of their generated storyboards.
Saved projects on the homepage
Easy access to previously edited videos.
Version control for edits
Undo, redo, snapshots. Fearless iteration.