<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Auboni Poddar on Medium]]></title>
        <description><![CDATA[Stories by Auboni Poddar on Medium]]></description>
        <link>https://medium.com/@aubonip?source=rss-5429b004ddcc------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*FwMZNxEB-wjmegZL</url>
            <title>Stories by Auboni Poddar on Medium</title>
            <link>https://medium.com/@aubonip?source=rss-5429b004ddcc------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 15:57:45 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@aubonip/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Media Controls: What We Learned from Testing Initial Prototypes]]></title>
            <link>https://medium.com/ai-widgets/media-controls-what-we-learned-from-testing-initial-prototypes-03f6ea778fe6?source=rss-5429b004ddcc------2</link>
            <guid isPermaLink="false">https://medium.com/p/03f6ea778fe6</guid>
            <category><![CDATA[interaction-design]]></category>
            <category><![CDATA[hci]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[ux]]></category>
            <category><![CDATA[carnegie-mellon]]></category>
            <dc:creator><![CDATA[Auboni Poddar]]></dc:creator>
            <pubDate>Wed, 11 Nov 2026 12:03:14 GMT</pubDate>
            <atom:updated>2026-11-11T12:03:14.683Z</atom:updated>
            <content:encoded><![CDATA[<p>Written by: Tvisha Arora &amp; Auboni Poddar</p><h4>Janky Prototype 1</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Ipb-q5hRosm_HJsu.png" /><figcaption>Prototype 1</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*RBnH3vAwA8Uim-11.png" /><figcaption>Audio Controls of Prototype 1</figcaption></figure><p>The first idea we’re exploring is a <strong>Customize Audio</strong> widget that uses AI to give viewers more control over what they hear while watching a video. In our prototype, users can select another video’s audio to play as background music while keeping separate playback and volume controls. The concept also uses AI to separate the original video’s audio into individual components, such as <strong>Instructor Voice</strong>, <strong>Original Background Music</strong>, and <strong>Exercise Cue/Timer Sound</strong>. Each component can be adjusted or muted independently. For example, someone following a workout could replace the original background music with a preferred song while keeping the instructor’s voice and exercise cues audible.</p><h4>Janky Prototype 2</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*dJTtOPDsrAdTGpL9.png" /><figcaption>Prototype 2</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/580/0*86FWIE6bYVS1wMhp.png" /><figcaption>Controls for Prototype 2</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mx2IyPefhRWLRQ5-.png" /><figcaption>Video divided into sections for Prototype 2</figcaption></figure><p>Prototype 2 explores how AI could make instructional fitness videos more adaptable in real time. The system divides the video into workout stages such as warm-up, individual exercises, recovery breaks, and cooldown, allowing it to understand where the viewer is in the workout. Users can <strong>Replay Last Instruction</strong> without manually rewinding, <strong>Restart Set</strong> from the beginning of the current exercise, or <strong>Extend Break</strong> when they need additional recovery time.</p><p>The prototype also includes <strong>Smart Audio</strong>, which adapts the music to the current stage and intensity of the workout. The BPM can change across warm-up, higher-intensity exercises, recovery periods, and cooldown. Users can also choose a music genre, adjust the balance between the instructor’s voice and background music, and decide whether they want the audio to feel more motivating or more relaxed.</p><h4>Testing Insights for Prototype 1</h4><p>Participants saw value in being able to separate and control different audio layers within a video, particularly the instructor’s voice, background music, and exercise cues. This would allow someone to lower an overly talkative instructor, increase the instructor’s volume when directions are difficult to hear, or replace the original background music while preserving useful audio from the video.</p><p>Some participants already mute workout videos and play their own music separately, which also removes the instructor’s voice and exercise cues. They saw Customize Audio as a way to personalize the music without losing those useful sounds. However, preferences around replacing the original music were mixed. Some participants preferred using their own playlists, while others liked the creator’s original music when it was already matched to the pace or intensity of the workout.</p><p>Participants also imagined applications beyond fitness videos, including recipes, tutorials, gaming footage, and other types of media where they might want to remove or replace one sound while preserving another. They also suggested adapting music automatically to different stages of an activity, such as using more energetic music during exercise and softer music during stretching.</p><p>One of the clearest themes from testing was that participants did not want audio customization to create more work. Manually adjusting sliders, changing songs, or repeatedly rebalancing audio during a workout could become disruptive. Participants suggested more adaptive behavior, such as remembering preferred volume levels, automatically balancing the instructor and music, or enabling voice-based controls. The walkthrough also revealed some usability issues, including audio controls being easy to miss and uncertainty around what certain volume controls affected.</p><h4>Testing Insights for Prototype 2</h4><p>Participants saw value in controls that could help them follow instructional videos at their own pace, especially when they fall behind, need additional recovery time, or want to repeat a specific step. Features such as <strong>Replay Last Instruction</strong>, <strong>Restart Set</strong>, and <strong>Extend Break</strong> were generally seen as useful for workouts, but participants also imagined using similar controls for cooking, drawing tutorials, and other step-by-step activities.</p><p>Replay Last Instruction was particularly appealing because finding a specific instruction with a traditional seek bar can require repeatedly rewinding and fast-forwarding. At the same time, some participants felt that a visual demonstration or preview of the upcoming exercise could be more helpful than simply hearing an instruction again.</p><p>Testing also revealed that the differences between some of the controls were not always immediately clear. Participants questioned how <strong>Extend Break</strong> differed from simply pausing the video and how <strong>Restart Set</strong> differed from replaying the last instruction. This suggests that the purpose of each action needs to be more clearly communicated.</p><p>Another major theme was interaction effort. Participants did not want to manage a large number of controls while actively exercising, cooking, or otherwise moving away from their device. Voice commands were repeatedly suggested as a more convenient way to replay an instruction, restart a set, or extend a break without interrupting the activity. Some icons also caused confusion because they resembled familiar actions such as shuffle or undo.</p><p>Participants were also interested in music that adapts to workout intensity, although preferences varied. Some wanted more control over the music themselves, while others preferred the system to automatically select music that fit the current stage of the activity.</p><h4>Summary / Next Steps</h4><p><strong>Next steps for Customize Audio:</strong> We want to expand Customize Audio beyond simply replacing background music and make the experience more adaptive with less manual interaction. Participants valued being able to preserve useful sounds, such as instructions or activity cues, while changing or adjusting other audio elements. They also identified possible applications beyond workout videos, including tutorials, recipes, gaming, and other types of media. Moving forward, we plan to explore automatic volume balancing, saved audio preferences, and context-aware music that adapts to different stages of an activity. We will also make the individual audio controls easier to find and understand.</p><p><strong>Next steps for Prototype 2:</strong> We want to simplify the number of controls and make the purpose of each action more distinct from existing playback controls. We will continue exploring features that help users keep up with instructional content, such as replaying an instruction, restarting a set, or extending a break, while also considering visual previews of upcoming steps. Because participants did not want to interrupt an activity to interact with the screen, we also plan to explore voice commands and other hands-free interactions that could make these controls easier to use while exercising, cooking, or moving away from the device.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=03f6ea778fe6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ai-widgets/media-controls-what-we-learned-from-testing-initial-prototypes-03f6ea778fe6">Media Controls: What We Learned from Testing Initial Prototypes</a> was originally published in <a href="https://medium.com/ai-widgets">AI Widgets</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Media Controls: A Closer Look at the Playback Timeline]]></title>
            <link>https://medium.com/ai-widgets/media-controls-a-closer-look-at-the-playback-timeline-418f86e47ad4?source=rss-5429b004ddcc------2</link>
            <guid isPermaLink="false">https://medium.com/p/418f86e47ad4</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[hci]]></category>
            <dc:creator><![CDATA[Auboni Poddar]]></dc:creator>
            <pubDate>Fri, 25 Sep 2026 22:38:49 GMT</pubDate>
            <atom:updated>2026-09-25T22:38:49.949Z</atom:updated>
            <content:encoded><![CDATA[<p>The playback timeline is one of the most familiar controls in digital media. Whether you’re listening to a song or searching for a scene on Netflix, the interaction is recognizable almost anywhere: a horizontal line representing the length of the media, with a playhead marking the current playback position.</p><p>I took a closer look at the media timeline, from learning about its background to seeing it in action on current platforms. I also explored a few research projects and observed five friends and family members using media timelines on their laptops to get an initial sense of how people use them and where AI might help.</p><h3>Background: How does the media timeline work?</h3><p>In <a href="https://doi.org/10.1145/3617448"><em>Pick, Click, Flick!</em></a>, Brad Myers explains that the media timeline fits into the history of scrolling and one-dimensional input. A scrollbar moves your view through content, while a slider lets you select a value within a range. A media timeline works as a slider, letting you select a playback position within a song or video.</p><p>The timeline shows where you are and lets you move somewhere else. But the same amount of movement can mean very different things depending on the media’s length. On timelines of the same width, a small movement along a two-hour movie’s timeline covers much more time than the same movement along a three-minute song’s timeline. This can make it hard to find an exact moment, especially if you don’t know where to look.</p><h3>What do media timelines communicate today?</h3><p>So, where are we today? I opened a few popular media platforms to see how they have built on the basic playback timeline.</p><p>On Spotify and Apple Music, I found the simple seek slider (“seek” meaning to move to a different playback position). While the timeline for music didn’t label sections such as the intro or chorus, Spotify offers synced lyrics for supported songs, which allow users to follow where they are in the song. <a href="https://support.spotify.com/us/article/lyrics/">Spotify lyrics</a>.</p><p>On YouTube and Netflix, I saw thumbnail previews that helped show what was happening at different points in a video. YouTube, however, stood out to me in particular because it offered several additional ways to explore content.</p><p>For example, YouTube offers chapters that divide videos into named sections. Creators can add these manually or allow YouTube to generate them, although they are not available on every video. YouTube also has an “Ask” feature on selected videos that lets viewers ask questions about the content. Together, these features offer additional ways to navigate and understand the video as it plays. <a href="https://support.google.com/youtube/answer/9884579?hl=en">YouTube chapters</a>, <a href="https://support.google.com/youtube/answer/14110396?hl=en">YouTube Ask</a>.</p><h3>What have researchers explored?</h3><p>I also took a look at three research projects that explored different ways to improve playback and navigation.</p><h3>SmartPlayer: adapting fast-forward speed</h3><p>In 2009, Cheng, Luo, Chen, and Chu developed SmartPlayer, a system that adjusts fast-forward speed based on video content and user preferences.</p><p>The initial study with 10 participants showed that people’s interest in fast-forwarding depended on the type of video. All 10 reported fast-forwarding surveillance footage, while none reported fast-forwarding movies. Similarly, 2 participants reported fast-forwarding lecture videos, while 9 reported fast-forwarding sports videos.</p><p>A later evaluation with 20 participants found that the system could reduce manual speed adjustments as it learned users’ preferences. Its usefulness varied across video types. <a href="https://doi.org/10.1145/1518701.1518823">SmartPlayer paper</a>.</p><p>What stood out to me most was that the reason a user is engaging with a certain kind of media may change what their needs are and guide which improvements would serve them best.</p><h3>RubySlippers: navigating through voice</h3><p>In 2021, Chang, Huh, and Kim developed RubySlippers, which lets people navigate how-to videos using voice commands that refer to the content.</p><p>In an evaluation with 12 participants, people completed navigation tasks with fewer commands and less frustration. The system explored how people could describe the part they wanted instead of having to know its timestamp. <a href="https://doi.org/10.1145/3411764.3445131">RubySlippers paper</a>.</p><h3>Content Flow Bar: previewing what comes next</h3><p>In 2022, Pérez-Ortiz and colleagues explored the Content Flow Bar, which provides previews of informational video fragments to help people understand how topics change throughout a video.</p><p>Participants found it helpful and enjoyable for finding relevant information, and the researchers reported promising signs that it could support exploration within and across videos. <a href="https://arxiv.org/abs/2201.03408">Content Flow Bar paper</a>.</p><h3>What I noticed in five observations</h3><p>To see how people use existing timelines, I observed five friends and family members finding moments in media on their laptops.</p><p>My participants using Spotify often clicked several positions on the timeline and listened to check whether they had found the part they wanted, while the people using video platforms hovered across the timeline and looked at thumbnail previews before choosing a position.</p><p>In my observations, people generally found their intended moments fairly quickly, even when the first selection was not exact. Spotify users listened before re-clicking and Netflix users looked at the preview before selecting, often getting close enough to what they were looking for.</p><p>These were only five informal observations, but they gave me a useful starting point. I want to understand how to distinguish what a pain point really is and what a solution looks like that doesn’t make this process harder.</p><h3>What’s next?</h3><p>This exploration left me thinking about the device someone is using, since my observations were done on laptops, the other ways in which people interact with media, and why people listen or interact with various types of media.</p><p>Someone following a tutorial may want to return to a specific step, but someone watching a movie may not need to see a breakdown at all. Someone listening to music may be trying to change their mood or accompany an activity. These purposes could lead to different needs, even when the basic timeline looks similar.</p><p>For my next step, I want to challenge myself to explore many solutions while keeping in mind the effort needed from the user compared with the current tools, so that a possible solution makes the experience easier.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=418f86e47ad4" width="1" height="1" alt=""><hr><p><a href="https://medium.com/ai-widgets/media-controls-a-closer-look-at-the-playback-timeline-418f86e47ad4">Media Controls: A Closer Look at the Playback Timeline</a> was originally published in <a href="https://medium.com/ai-widgets">AI Widgets</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>