Project Overview

Noise Eraser APP & Website - 2022.01 — 2023.02

Product background

Noise Eraser is an AI-powered noise reduction tool designed for content creators and remote professionals. It delivers fast, hassle-free audio cleanup without the need for manual editing. The product’s core vision is to make noise reduction feel effortless yet trustworthy—addressing common challenges in traditional noise reduction tools such as complex controls and inconsistent results.

01 | Team Goals

  • Increase user engagement and improve membership conversion rates.
  • Reduce user churn and shorten the learning curve.
  • Ensure UI/UX consistency between the app and web versions to enhance the overall user experience.

02 | My Role & Deliverables

Served as the sole designer, designing all UI/UX (wireframe, user flow, UI mockup, prototype) initiatives in collaboration with one PM, one Android/iOS engineer, and one front-end developer.

03 | Project Challenges

In this AI noise reduction project, our key challenge was: How do we build trust in a system that works instantly yet invisibly?
Unlike traditional manual editing, AI denoising is fast but opaque—users couldn’t tell if it worked, leading to confusion and doubt.

  • New users often couldn’t tell if the noise reduction was working at all, leading to confusion and mistrust.
  • Advanced users wanted more control and clarity without sacrificing the speed and simplicity promised by AI.
  • System-recommended outputs didn’t always match users’ creative intent, such as preserving room tone or vocal warmth.

This project aimed to make invisible AI actions visible and understandable—so users could trust and confidently share their results.

04 | Outcome

↑ 35% Avg. session duration

↓ 22% Bounce rate

↑ 0.61% Total subscription count

↑ 27%  Returning user rate

User Feedback

Apple Store
Google Play

Problem Discovery

After the product launch, the team aims to enhance it further and gain a better understanding of our users' profiles. At the COO’s recommendation, we initiated a user research phase that included persona development to guide marketing strategies. We also conducted user interviews and usability testing to identify usability issues and inform product iteration.

Research Method:
I was responsible for formulating the interview questions and conducting sessions with six users to explore their behaviors and pain points around noise reduction tools. Collaborating with the PM and user researcher, I compiled the findings, and together we mapped the user journey and developed personas during a workshop.

Key Findings:
1. 66% of users churned due to the inability to clearly perceive the noise reduction effect.
Since the product is new to the market, users are still exploring its functionality. Some videos naturally have minimal noise, making the effects less noticeable.
Without a clear way to demonstrate the feature’s impact, users might misunderstand its effectiveness.
2. 50% of users felt confused due to differences in app and web interfaces.
3. 66% of user are not sure whether our product has auto save function.
4. 33% of user think that the outcome is not aligned with the user’s expectation

Problems

We also conducted a usability test and user research, during which I was responsible for interviewing users. Based on the findings, I summarized the following four key issues:

1. Users are unsure about the usage status → Lack of real-time feedback leads to a gap in user experience.

2. Inconsistent UI/UX between the app and web versions → Increases the learning curve and affects user retention.

3.Unclear file-saving status → Creates uncertainty, disrupting the workflow.
4. The outcome is not aligned with the user’s expectation→ Reduces productivity and lowers conversion rates.

1. Unclear state
Users are not sure if the post-noise reduction video is the output of noise reduction or the original video.

1.2. Noise Reduction Feature Analysis

Noise reduction effectiveness varies depending on the video. Some files show little difference before and after processing, making it hard for users to perceive its impact.
This may lead users to believe that the product is ineffective, potentially resulting in user loss.

1.3.Voice and Background Sound Adjustment

If both voice and background sound are set to the same value, no reduction occurs, which may further reinforce the perception that the product is not working as expected.

Since the product is new to the market, users are still exploring its functionality. Some videos naturally have minimal noise, making the effects less noticeable.
Without a clear way to demonstrate the feature’s impact, users might misunderstand its effectiveness.

2. Inconsistency in the operation methods between the WEB and APP

Users said if a large number of files need to be downloaded, it is preferable to rename the files during the editing process for easier organization after downloading. Renaming the files only at the time of downloading would require individually checking each file, which is more cumbersome.

Currently, it can only be modified on the APP.

Inconsistent default icons and premium badges (crown icon) between the APP and the Web.

3. Missing basic information and functionality

While users edit the file, they are not sure whether the file will be saved with the currently adjusted background sound and voice ratio or not.

4. The outcome is not align with user’s expectation

When users complete adjusting the audio and share the link with others, they anticipate the shared page will reflect their real-time noise reduction settings.

However, instead of displaying the user's specific adjustments, we provide a before-and-after comparison showcasing our noise reduction capabilities. This approach allows us to highlight our expertise in noise reduction on this webpage.

When sharing a link, the default setting of the dropdown menu is set up in "private" which doesn't align with the mindset of users who already want to share it with others.

Problem Statement

Users often feel uncertain or anxious when sharing meeting content automatically. They lack control over what is shared, when, and to whom. This causes hesitation and may reduce engagement with automated productivity tools.

Our goal is to design a more transparent, customizable, and trustworthy sharing experience that aligns with users’ mental models and boosts adoption.

Solutions

This redesign focused on three main goals:
(1) Reduce ambiguity in how the denoising works,
(2) Offer users both default simplicity and advanced control,
(3) Build user confidence in the system through clear status feedback and shareability.

1. Enhance real-time noise reduction visualization

Remove the play button from the thumbnail.Instead, use clear status labels such as “Processing,” “Edit,” or “Complete.”

  • Design Rationale:
    I observed that the presence of a play icon suggested the file was ready for preview, creating false expectations. In reality, the noise reduction might still be processing or only partially complete. Removing the play button ensures the UI reflects the actual system state and guides users more accurately.
  • Hypothesis:
    If I remove the misleading play icon and replace it with explicit status labels, users will:
  • Better understand the current file state.
  • Feel more confident that the noise reduction is working.
  • Be more likely to proceed with the next step (edit, download) without hesitation.
  • Trade-off:
    I considered keeping the play button to allow users to preview audio directly. However:
  • It risked playing the original, unprocessed audio, which could cause users to think the tool failed.
  • Clarifying this distinction through tooltips or dialogs added cognitive load and interrupted flow.

We prioritized clarity and trust over offering early playback interaction.

Noise Control: Combining Presets with Custom Flexibility

  • Design Rationale:
    A gray but draggable slider allows users to see the preset noise reduction levels (e.g., voice vs. background mix) while maintaining control to fine-tune. This approach balances transparency and customizability, giving users clarity on system logic and freedom to personalize.
  • Hypothesis:
    Users will better understand how each preset works and feel more in control when they can adjust values directly. This will lead to increased trust in the product and reduced frustration when outcomes differ from expectations.
  • Trade-off:
    The gray appearance may initially suggest the slider is disabled, potentially causing momentary hesitation. However, I offset this by ensuring instant visual feedback—once the slider is touched, the label switches to "Custom," reinforcing interactivity and system response.
2. Unify the app and web user experience. Reduce the learning curve and increase user retention rate.

Add the edit function of the file name.

Make the prime icon the same between different platforms.

I refer to Canva and Google Slides which are popular editing tools, and I found that they use tooltips to help users understand the icon's function.

To decrease the user's learning curve, I follow their solution. When users hover over the background and voice icon, a tooltip notification appears above the icons on the web. Moreover, as for the mobile version,  it will show the tooltip for several seconds when the users jump to this page.

When users click the different buttons, a tooltip notification appears on the video to inform the user of the currently used filter.

Validation:

This solution aligns with established UX best practices (e.g., Nielsen Norman Group’s heuristics).

3. Add real-time autosave notifications (Autosave Toast) → Reducing operational uncertainty.
  • Design Rationale:
    Inspired by tools like Google Docs and Canva, we introduced a real-time autosave toast that appears as soon as editing begins. This lightweight yet visible feedback mechanism is designed to reduce operational uncertainty, especially for users unfamiliar with the system. Unlike mature tools (e.g., Figma) that rely on user trust, our early-stage product needed explicit, reassuring system feedback to build confidence from the start.
  • Hypothesis:
    I believe that by clearly communicating autosave status in real time, users—especially new ones—will:
  • Feel safer and more confident making edits
  • Form an accurate mental model of how the system handles saving
  • Avoid confusion or anxiety over potential data loss
  • Trade-off:
    Compared to a silent autosave model (like Figma), visible toast notifications introduce more UI surface noise and may appear redundant to advanced users. However, for our primary audience—many of whom are new to the tool—clarity outweighs minimalism. I mitigated interruption by keeping the toast brief, non-blocking, and positioned in a low-visibility zone.
4. Balancing Brand Showcase with User Needs: Optimal vs. Custom Denoising Ratios

"In testing, 4 out of 5 users said they preferred sharing their custom result over the system-recommended one."

  • Design Rationale:
    To balance brand consistency with user expression, I introduced a toggle feature that allows users to switch between the system-recommended “Optimal” denoising ratio and their personally adjusted “Custom” settings. This design supports two different priorities:
  • The company’s goal of promoting algorithmic excellence through the cleanest output.
  • The user’s desire to share a version that aligns with their creative intent—be it preserving ambient nuance or vocal warmth.
    By clearly displaying the active ratio and offering switching flexibility, I provide both transparency and autonomy.
  • Hypothesis:
    I believe that giving users control over what version gets shared will:
  • Increase their trust in the system and willingness to share outputs publicly
  • Reduce dissatisfaction caused by mismatched expectations
  • Strengthen the product’s credibility among creators who value customization
  • Trade-off:
    Compared to forcing a fixed “Optimal” output for all shares, this toggle introduces a more complex UI decision point.
    However, I found that:
  • The toggle is simple and lightweight in design
  • Users feel more empowered and in control
  • Technical implementation requires no algorithmic recalculation, making it a low-risk addition
    I considered an alternative that only allowed “Optimal” sharing, but it limited personal agency and backfired during testing.

Future Plan

Although key usage metrics showed positive growth post-launch, overall revenue remains below expectations. To strengthen monetization and capture long-term value, I proposed the following future initiatives:

1. Introduce Tiered Pricing & Freemium Model

  • Why: Users are hesitant to pay without trying. A free tier lowers friction.
  • Plan: Launch a limited free version (e.g., watermark, 3-minute audio cap), with Pro and Studio tiers offering higher limits and batch processing.
  • Expected Impact: Higher conversion from free users and improved lifetime value.

2. In-App Upsell Touchpoints

  • Why: Users often realize value during usage, not before.
  • Plan: Embed upgrade prompts at key moments (e.g., after previewing processed audio or during batch exports).
  • Expected Impact: Monetize emotional “aha” moments and increase upgrade rate.

3. Launch Team/Business Plan (B2B Use Case)

  • Why: Studios, podcasts, and agencies often need multi-seat plans.
  • Plan: Offer volume licensing with admin features and cloud project sync.
  • Expected Impact: Expand revenue channels and increase ARPU through B2B.

Key Metrics to Track

  • Free-to-paid conversion rate
  • Upsell engagement rate
  • ARPU (Average Revenue Per User)

Takeaway

This project strengthened my ability to translate complex AI technology into intuitive, benefit-driven experiences for both end users and business stakeholders. Beyond interface design, it became a deep exercise in cross-functional collaboration and design strategy.

  • Making Progress Visible to Build Trust
    I learned how critical system visibility is during tasks like audio processing. By making progress and results more transparent, I reduced user anxiety and improved clarity, trust, and perceived control.
  • Base Design Decisions on Business Impact
    Good design isn’t just visual—it must align with business goals, technical constraints, and timeline realities to be truly effective.
  • Bring Engineers Into the Research Phase Early
    Early involvement of engineers helped bridge the gap between concept and feasibility, building stronger buy-in and smoother execution.
  • Be Open to Being Wrong
    Admitting uncertainty welcomed diverse perspectives and enabled better design outcomes through feedback and iteration.
  • Communicate Clearly and Respectfully
    I focused on balancing user needs with business objectives, fostering productive, trust-based collaboration.
  • Always Assume There’s Room for Improvement
    Maintaining a humble mindset helped me refine the design continuously and remain receptive to team input.
  • Use Loss Aversion When Presenting Ideas
    I found that framing poor UX as a business risk—not just a usability issue—was more persuasive when aligning with stakeholders.

Through this project, I evolved not only as a designer, but also as a communicator and problem-solver—able to bridge technical, business, and human needs through thoughtful UX.

Related Products

RecruitmentApp

UXDesign

JobPlatform

TalentMatching

MobileUX

B2BDesign

HR Tech

FormOptimization

Web3Design

CryptoUX

FintechDesign

Blockchain

MobileUX

BlockchainUX

B2CDesign

DeFiDesign

UserManagement

WebUX

PlatformUX

Account Management

EnterpriseUX

Admin Dashboard

Welcome to contact me !