Final Design
Screenshot | Explanation |
|---|---|
| Originally, the main screen looked like the extended song list below. Our final design adopts a splash-screen-like design in order to highlight the name of the app and reassure users that they're in the right place, while remaining functional with a prominent "new song" button. The background is intended to be aesthetically pleasing but unobtrusive. |
| We limited the main screen to three saved songs to avoid overwhelming users. The "new song" and "about" buttons remain prominent. Tapping on a song launches the compose screen, shown below. Our original design included small "rename", "duplicate", "export", and "delete" buttons on each song, but our paper and heuristic evaluations noted that such nested buttons were confusing and difficult to use on a small screen. We therefore replaced them with a menu accessed by tapping and holding for a second or so; this method is nonobvious, so we also included an instructional note. |
| Renaming a song works by replacing the name in the song button with an editable text box; when the user submits (presses enter) the song is saved. Duplicating a song creates a new song named "Copy of NAME" and enters rename mode on that song. Deleting a song asks for confirmation, then slides the song button rightward off the screen. (It would perhaps be clearer for the button to shrink to nothing vertically with the buttons below sliding up, but technical properties of the Android API make implementing this animation difficult.) |
| Exporting uses native Android sharing functionality to display a list of possible transfer methods for the user to pick from. A MIDI file containing the song is sent via the specified method. |
| If the user clicks the "see all" button, this screen is shown. It lists all saved songs, allowing scrolling and searching (the list updates as a user types in the search box). |
| The "about" button takes users to this screen, which is just a static message. |
| This is the compose screen for a song with no notes in it (e.g. a newly created song). Swiping horizontally scrolls the song and two-finger pinching zooms it, as is common for touch interfaces. |
| A click track plays while the user is recording. The cursor turns the color of the current track, and moves rightward across the screen (the display scrolls so that the bar is never more than 3/4 of the way across). Notes are created at the cursor position when the app detects singing, and lengthen until they change pitch or stop; notes are deleted if the cursor hits them while recording into the same track. All buttons are disabled here, but "Stop" is available to stop recording. |
| Here the user has stopped recording on the green track, and started recording on the purple track. All four tracks are drawn on top of each other; notes on the same track can't overlap, but notes on separate tracks can. |
| Here notes have been recorded into all four tracks. Notes from the currently selected track are bright and large; other notes are dim and small. |
| The 'new' editing interface. Tapping on a note causes these icons to appear. Tapping the trash can deletes the note. Dragging the side arrows left or right changes the start and end points. Dragging the middle icon vertically changes the pitch of the note, and dragging it horizontally moves the whole note without changing its length. As notes in the same track can't overlap, the note stops when it hits something, and a red glowy thing appears to signal that it can't move or extend anymore. |
| The 'old' editing interface. Tapping on a note causes this overlay to appear. Tapping the trash can deletes the note. Swiping left or right on the left side of the screen moves the start of the note backward and forward in time; swiping horizontally on the right side does the same for the end of the note. Swiping up or down in the middle of the screen changes the pitch. There's no way to move the note without changing its length. If the note tries to expand but hits something, it stops and a red glowy thing appears; also, the arrows that don't do anything are crossed out. |
| Tapping on the track tools button brings the user to this screen, which has color-coded controls for each track. A muted track doesn't appear in the compose screen and produces no sound during playback. |
| The instrument menu allows the user to choose what musical instrument the track will be played as. These selections are saved with the song, so they persist across editing sessions. |
| If there are unsaved changes to the song when the user tries to leave the compose screen, the app prompts the user to save. There are three options: save then exit, exit, or don't exit. |
| When saving a newly created song, the user is prompted for a song name. After that, saving uses the existing name, and renaming from the song menu is necessary to change it. |
Implementation
Most of our app is written in Java using the Android APIs, but there are native libraries to perform playback and recording. (We used native code because there were no convenient Java libraries, and implementing the functionality ourselves would have been difficult and possibly resulted in slow and buggy code. We used a pure Java / Android API implementation for playback in the prototype, and the quality was fairly terrible.) There is also a third-party library (com.leff.midi) that implements the MIDI message protocol and file format. Recording grabs audio samples from the microphone and outputs pitches, from which the compose code can create notes. Playback takes notes and sends MIDI messages to a system audio library, which produces audio output.
Beyond those, the only important backend class is Song, which encapsulates the concept of a song with a name, a mapping of tracks to instruments, and a set of notes, each of which has a track, pitch, a start time, and a duration. Songs can be written to and read from file storage; the app creates a file in private storage for each saved song. Songs can also export themselves to MIDI files.
An activity is an Android concept described as "a single, focused thing that the user can do". It's typically a single screen or view that the user can see and interact with. Our app has four activities: the main screen, the about screen, the full song list screen, and the compose screen. The main and song list activities contain SongWidgets, which are the buttons in the lists; each one corresponds to a song and has code for interacting with that song.
The compose activity implements all the various button actions, interacting with recording, playback, and file storage. It has various state including cursor position, selected track, and zoom level. It also contains a list of notes. Each note is a custom GUI widget that draws itself onto the compose screen; it keeps track of its track, pitch, start time, and duration, and adjusts its color, position, and size as appropriate.
Each note listens for touch events; if the note is tapped, it starts the edit mode. In the 'old' interface, this is a full-screen semi-transparent widget that detects swipe events and modifies the note based on their starting point, direction, and length. In the 'new' interface, this makes the various icons visible; they listen for touch events and modify the note when dragged.
The compose activity contains the undo stack, which is a list of UndoItems. An UndoItem is added whenever a change is made, and the stack is modified when the undo and redo buttons are pressed. When the user saves a song, the compose activity converts its internal data structures into a Song, and saves it to file storage.
We planned to play back notes in other tracks while recording, à la overdubbing, so that users could more easily create harmonies when recording different tracks, but the microphone picked up the notes and it messed up the pitch detection. We considered something like disabling that feature unless headphones were plugged in, but concluded it wasn't worth it.
Evaluation
Our users were friends from our living groups who had varying levels of interest in music. While they're all MIT students and so might not be generally representative, MIT students seem to have a pretty wide range of musical inclination and so weren't a particularly bad sampling of our user population, though most of them were programmers (don't think any of them had programmed on Android before, though). Our app aims to be usable by people with varying levels of music experience and who are probably somewhat younger, creative and tech-savvy enough to carry a smart phone.
We conducted user tasks by telling the user the purpose of the app (record songs without having to figure out the musical notation and when you're on the go) and the target audience (should be usable by people without a lot of musical experience).
We mentioned that they should think out loud, and that we were testing the app's usability, not their intelligence.
We had two editing interfaces that we can't decide between. We tested this by having different users test different editing interfaces first, give us their comments on it, then switch and give us their comments on the other one. I should've been careful to not tell users who made which interface, but I think I failed at this for the first two.
We tried to stay silent during observation, only giving tasks, then at the end asking questions about certain things the user did that they didn't explain, or things they may not have discovered.
Tasks
1. Record something
2. Edit
3. Add a harmony line
4. Export
User summaries
User 1
Starting with (old edit) interface
Course 6
Android User
Experienced Chorollaries singer
User 2
Using (new-edit) interface first.
Course 6
iPhone user
Casual composer, doesn't often finish songs
User 3
Starting with (old-edit) interface.
Singing and playing instruments since he was 7
Doesn't own a smartphone, so I briefed him on Android usage. Function buttons, long tap, and zoom gesture.
"I'm definitely not a computer person"
We are sitting next to the piano and I mention that piano records nicely.
Bit worried about this user test because he seemed unwilling to hurt our feelings or worried about not making mistakes.
User 4
Starting with (new-edit) interface
"I warn you, I do usability testing for my lab, so I will, like, give you user interface feedback."
singing for as long as can remember in a wide variety of genres
"I just finished singing, so I'm warmed up"
Course 6
"I don't have any theory background, only performance."
Discovered usability issues
The "Users" column indicates how many (out of 4) of our testers had that issue.
Issue | Users | Possible fix | Severity |
|---|---|---|---|
Small notes are too difficult to tap; caused some cursing | 4 | We added a "hit extender" to help resolve this, but it doesn't help in the x direction. We could potentially add next/prev note buttons. | Critical |
Tries tapping arrows in old-edit rather than swiping them | 4 | Usually they figure out how to use it fine after trying a few more things. | Minor |
Accidentally start recording/play in the wrong place | 4 | Maybe copy Finale's method of playing from beginning usually, not sure about record. Pretty recoverable. | Minor |
App crashes after export | 4 | Bug | Critical |
Phone doesn't pick up melody anywhere near accurately | 3 | ?? Editing or re-recording is often sufficient. User often needs to hold mike closer to mouth or sing with a steadier voice (vibrato does not seem to help), maybe if they have to undo recording several times the program can suggest this. Recording seems to record just as well with words being sung as with random syllable, so it's not that. A user suggests that the app display your actual voice pitches so they can tune the lines to the voice. A user resorts to using piano to record, but this isn't our use case. | Major |
Instrument change/track muting not visible | 3 | 1 person only knew where it was because he found it while looking for something else. These features need to be more visible. One user suggests double tapping track button to bring up options. | Major |
(old-edit) should have move left/right function. 2 actions is "bizarre and awkward" | 3 | Main/deciding reason people cited for preferring (new-edit), wouldn't be hard to just add. | Critical |
Confusing back button with okay in old-edit interface, even after using many times | 3 | Perhaps move them closer together, where you can see them all? | Major |
If a song contains two notes on the same pitch, they often fuse, and there is no way to split notes | 3 | Add a split note function | Major |
Feels like it's impossible to coordinate harmony with melody | 3 | 1 was just really careful, 2 suggest playing the melody back while recording harmony, 1 suggests program do some AI to coordinate melodies and note lengths. | Major |
(old-edit) Arrows obstruct note you're editing | 3 | This is especially common on notes near start of song, since they can't be centered. Maybe if overlay were more transparent. | Major |
Want to move/delete segment of notes en-masse | 2 | a third person had an error in their music that would've been much more easily resolved with this function. We should have it. | Major |
Wants to be able to have notes same length/multiples of same length like in sheet music | 2 | Maybe we should do some AI for this? | Major |
It's not obvious that notes are editable | Unclear, at least 3 non-tested people | ?? | Critical |
Accidentally presses play when they mean record. May not realize it even after trying several times to record. | 2 | Not sure what's causing this? | Minor |
Found export difficult to find | 2 | ?? | Major? |
Stop is difficult to find while recording. | 2 | Seems like users expect tapping the cursor to stop the song, or that stop should be red. We might try changing that. | Minor |
Can't tell which buttons are disabled | 2 | Might be more but I only asked 2. Should make buttons non-gray or something | Minor |
Phone records an octave above where it should, only very briefly though | 2 | This and other noise during recording are easily deleted in interface (Besides the fact that they're too short) | Cosmetic |
Didn't notice undo button | 2 | ?? | Minor |
Accidentally go into edit mode for a note when trying to move cursor | 2 | ?? | Cosmetic |
Wants to zoom closer/farther than min/max zoom | 2 | Minor | |
Didn't notice zooming into scale was possible | 2 | Might be more than 2, but 1 was instructed. ?? | Minor |
(new-edit) Accidentally presses trashcan while editing | 1 | Moved trash can after this user | Fixed |
Wants a next/prev note button when editing | 1 | Only 1 mentioned this, but others were also editing essentially linearly | Cosmetic |
Don't realize what note length not being able to increase means | 1 | This was fixed after a user test by implementing red bars when notes butt against each other. | Fixed |
Does not realize that tracks record other overlayed melodies | 1 | He actually recorded over his original melody, which was pretty bad. Saw his grayed-out melody on other tracks and thought they were other people's songs. This came up in heuristic testing too; it's unclear that people realize what tracks are until instructed to "record a harmony" | Major |
Play doesn't begin from where expected | 1 | Cursor did not follow editing, and played from seemingly random place. Cursor usually moves to thing you're editing in word editing, so maybe it should. | Minor |
Wants to be able to add notes | 1 | We could add this somehow | Cosmetic |
(old-edit) Loses pitch while trying to change note and has to keep re-playing | 1 | User suggested this would be fixed with next/prev note function. Another user hummed the pitch they wanted while adjusting to keep track of it. | Minor |
Note disappears while editing (new-edit) | 1 | Bug | Critical |
Buttons at top of screen make user cover screen | 1 | switch top and bottom buttons | Minor |
Wanted to hold record to record | 1 | ?? | ?? |
Wanted to be able to pull notes past each other | 1 | We could make notes on a track switchable. | Cosmetic |
The default instrument does not sustain, doesn't match interface | 1 | We could switch it | Cosmetic |
(new-edit) wanted the delete button closer to note | 1 | Conflicts with another user's accidentally pressing trashcan | ?? |
During playback, single note on another track played twice (replayed when a 2nd note on another track started) | 1 | Bug | Minor |
Looked for exported song in my files | 1 | He was one of 2 actual android users, maybe other users would've tried this. Not obvious where to look for export. | Major |
Feels like scroll is broken when recording is shorter than phone length | 1 | Came up in a heuristic test too. Maybe make scrolling available anyway | Minor |
(old-edit) Shortening/lengthening note was not working, note wasn't even visible on screen | 1 | Bug? | Major |
Reflection
"Our project was so shiny we used it as a mirror." - David Benjamin
"Despite the limitations of paper prototyping in representing our application, we learned that the spiral model of development was quite valuable, as our project saw a pretty large overhaul between user testings during that iteration of design. We learned that what users expect can often be pretty unintuitive; for example, there are a lot of expectations about where buttons are and how they will work that users probably couldn't tell you they expect to see, but that their behavior in repeatedly using the wrong button shows (e.g. using play as record and thinking it's broken). Personally, I learned some Java and Git from Alan; I hope I didn't frustrate him too much. If we had to do it again, we should probably have gotten GR4 in on time so that our testers didn't get reassigned. That was kind of unfortunate." - Di Liu
"We prototyped the easy things and not the hard things." - Alan Huang
"We evaluated the results of our observations by, I don't know, thinking about them." - Alan Huang
"Really hard." - Alan Huang
"Totally." - Alan Huang
It was fun! Thanks!
View Source
User Testing
Tasks
1. Record something
2. Edit
3. Add a harmony line
4. Export
Transcripts
User 1
Android User
A capella singer
Starting with old edit interface, then getting commentary on new
1. Record Something
- Records without looking at phone, trusting it to work
- "Is it gonna play back as voice or what?"
- He sings a fast song kind of quietly and the recording is um, is quite different from what he said
- Uses undo to try again.
- "twinkle twinkle..." Sings with the words
- It does better but gets some noise and merges many double notes
- "It has a problem with notes on the same pitch, but I would probably just articulate more if I wanted it to work. That might be hard to actually do."
2. Edit a note
- "How does one edit notes?"
- "Wah!
- Drags pitch
- "Wait, did that even move?" Arrows might be blocking note
- plays to confirm
- "Wait, stop!" Seems to have a bit of difficulty finding stop, but it's the first time he used play after all.
- "I wanted this note!" and moves it to the right place, but clicks back.
- "Oh check box! I didn't see that. I pressed the back button thinking it was okay."
- "Can I split notes?"
- Tries various buttons to see if he can find split
- Trying to use arrows that have x's over them, probably mostly looking at the note he wants to edit and stroking based on that
- Tries left/right stroking.
- "that made it shorter..."
- "Gah, not clear that that's possible"
- accidentally clicked a note trying to move cursor
- "I like the swipe up and down to get the right pitch, cuz you can hear what you want."
3. Record a harmony or something
- "Let's do it in a round."
- accidentally started recording at end rather than start
- accidentally pressed play to record, then surprised no notes show up
- gets right button
- this time looks at screen to record, sometimes. Maybe trying to line it up but gives up
- seems to figure out how tracks work just fine
- "at the beginning I was trying to track it... but it's completely off."
- "now it will sound completely weird"
- presses play
- it does
- "it's hard to record harmony without knowing the original tune."
- "Now that's modern music!"
- "I'd definitely want to hear what I'm recording against, but I don't know how easy that is"
- "you could definitely recommend using headphones, but it's really hard to record more than one track without that."
4. Export song
- "Okay I'm gonna save it as."
- "Uh, Modern twinkle twinkle. save."
- back
- "Tap and hold for options. Ok, let's export it to midi!"
- "whoah I can send it to (all these things)!"
- "Where does wifi go? Does it just go, randomly somewhere over the network? Oh, so it looks like wifi direct. Ok I don't know what wifi direct is."
- "I'll email it to myself! Look I'm in alan's contact list."
- "Now let's ask myself if I got a midi! Well, not yet."
- App crashed awhile after export.
Nudge questions:
- Did you notice you can change instruments?
Wiki Markup "No. Oh, I didn't try this \[settings button\]. Aha\! All of these guys look familiar"
What did you think of note editing interface?
- swiping all directions
- "I like the up and down, that's for sure. The moving the note left to right seems sort of difficult. For one it seems to move in big chunks. The x's mean you can't move it any further, right? Obviously you can't move a note to zero. I guess you have a maximum note length? That doesn't seem so reasonable. Or it just doesn't, wait, I guess it's not clear to me whether that's a max note length or, oh, I see, it goes until it butts the next note in the same voice. That makes mores sense. Yeah. I guess it's hard to deal with, like say I wanted to move a note forward a little bit in time, that would be kinda difficult with this interface. At the same time it is also useful to stretch and shrink notes, I think it's hard to get all of that in there."
Try (DD's) interface. I tell him my phone takes longer to record things.
- Means to hit record
- "Oh that was play."
- "hellooooooo helllooooooooooo!"
- seems to record okay, gets random noise an octave above.
- "awo!"
- has a bit of a hard time pressing the note.
- "it doesn't have this whole save making changes to a note, or stop making changes to a note, you just click away. I see."
- "It's a liiiitle strange. You have to click directly on a note and then drag it to move it, whereas in alan's you can click anywhere to drag it. What's the most confusing about this up and down part is that I have to drag it, but it doesn't follow my finger exactly. Okay let's try left and right."
- "Okay so that moves it left and right, oh I like that."
- "and that makes it bigger, okay"
- "I like the left and right. I definitely like the left and right better. You can do that without changing the width of the note, and you can change the width. That's cool, sweet."
- "It does seem to be easier to do a lot of different notes."
- "I definitely noticed on Alan's you can just click undo if you don't want your change... But since you have undo up here, that's probably..."
Compare them.
- I think that.. hmm. (thinks for quite awhile)
Wiki Markup (Playing with alan's) "is this actually centered on the note? hmm I think I would kind of like Alan's better except with left and right functionality more like yours. You know with move \[in all directions\] in the middle" (lots of gesticulation)- "it's true getting from note to note is much faster in this one (DD's)"
User 2
Using DD's interface first.
iPhone user
Casual composer, doesn't often finish songs
1. Record a song
- Clicks on a bunch a buttons
- Hit record
- "okay alright...."
- Sings "twinkle twinkle", except with doooooos, looking at it.
- Phone picks up pitches, kind of.
2. Edit
- taps a note. Trying different things on it.
- "So what I wanted to do was split a note in half. I guess that might be part of your phone thing but yeah."
- tries to tap end arrows with no effect
- deletes note
- "I was trying to click note between those notes (a row of 3 min length notes) and then I tried to tap away and then I clicked another note and it was hard to get away because all these things appeared and all I wanted was that note."
- "I tried to play, but I meant play from start." (uses to start)
- "I don't know if to start without a just play from start makes sense. I guess you could use it for record."
- plays back and stops to edit things. Uses dragging left to right a lot. Change pitch very carefully by tapping on note.
- "Is there like a go to next note button?"
- "It's hard to drag the pitch of small notes."
- Notices from red edge that she needs to shorten next note to lengthen this.
- "It's really hard to click small notes."
- "oops. Stoop!" Has hard time finding/hitting stop. Taps everywhere.
- "I was totally editing a note and I pressed play and it jumped to a nother place, what was up with that?" (cursor was in another place. Maybe cursor should follow editing)
- "wait the cursor isn't on my note?"
- doesn't notice you can stroke anywhere in the vertical
- "oh shoot I deleted. Undo. Yeaaah"
- "Darn I want another note. Oh look another note!" Take little extra note and drags it over
- "I really like how you can just drag a note around"
- "It turns out it's really easy to hit the trashcan cuz sometimes you have notes right over the trashcan you wanna hit. Esp if you have a descending line."
- "I think my song gets faster. (as in the lengths get shorter at end.) I really wish I could extend a note and have all notes move over, or have a mass move button, rather than moving them all one at a time."
- "I keep meaning to play from start but it just plays from whereever."
- finishes
- "Yeah! alright."
3. Record another track
- "...I don't know any harmonies to twinkle little star."
- (presses play rather than record again)
- (Singing) "pleeeeease record my song!"
- After several tries of clicking play, realizes it's not record.
- Realizes record doesn't do playback
- "Awh! I don't get to hear my song now? how am I supposed to harmonize with myself!"
- "Damn I can't click small notes!!"
- "Agh I just dragged my note to the side and it disappeared."
- bug. Various terrible things happen.
- she re-records
- Noise in the octave (she sings one note but it alternates between octaves.
- "I'm not sure how it got those ones"
- "wait can I no longer zoom in?" (she's at max zoom"
4. Export the song
- "Too bad, I'm naming it."
- "the star tried to twiiiiiiiiinkle. done. save. Yeah!"
- settings: "What's this? Oh cool!"
- "Wait how do I get outta here? ahhhh!" (not an android user)
- giggles while playing track in helicopter
- "wait what was I supposed to do now?" (re-instructed)
- "tap and hold for more options."
- chooses export.
- it crashes.
What do you think of this editing interface?
- "I really like how you can just move notes around. The trashcan kept getting in the way. I think it'd be nice to have a next/prev note function instead of playing the whole thing cuz I know where the pitches are sposed to be to hear what's wrong. And it's extremely hard to click small notes, holy shit."
Try Alan's interface
- Katfang tries to click play button to record again, despite already making this mistake and it being disabled.
- "Oh my god. it's different."
- "oops I pressed back instead of check"
- "Oh my god I have to click small notes again. Did I do it? I can't tell!" (arrows covering note, seems to be a problem with early notes)
- "I gave up trying to remember what the right note was while editing. I want to go one note at a time, cuz I don't actually know what note (pitch) I want, but I know the interval, but there's no next note and you have to go out every time."
- Still tries to tap arrows to change note even though she knows it's swipe
- About to use back
- "Oh shoot no. Check."
- "I want to cut out the beginning part, where there are no notes, cuz I'm like what the hell am I doing k I'll start now."
- "I'm still trying to tap these arrows even though I know they don't do anything unless I swipe over them." (shortening and lengthening)
- keeps trying to press back instead of confirm.
- "Stop stoooop!" tapping everywhere. "Oh stop."
- "Moving notes is bizarre and awkward." (messing with arrows to scoot notes over)
- "Nope": tries to play from wrong place agin.
- (trying to get note lengths right) "One thing that kind of bugs me is that I can't tell how long notes are relative to each other. Like a quarter note or something. Like if one is twice as long as another. I can only tell how long a note is within 5 seconds."
- "Also I have no music sense so I keep having to replay." (loses the note she wants while editing because it plays)
- "what is this note supposed to be??" (tries many times to get pitch)
- "Darn this was the actual note it recorded and that was just a messup note. Darn! Darn! I'm so angry!"
Impressions on this interface
- "My biggest thing was that I couldn't move notes. I just want to take this chunk of notes and be like you move that way! There's this big gap and I don't want it. I didn't randomly delete notes like the other one.
- I had a really hard time remembering the back button vs confirm, I kept clickng back note."
- "This interface especially could have a next and prev note. I dunno if that's good, if that's too many buttons. I kept having to click a note and pressing back and playing and clicking a note and pressing back and playing. You know cuz intervals. This would totally be fixed with a next/prev button."
- "turns out it's really hard to type potter. Oh well he'll just have to be pretty."
Compare interfaces.
- "I liked this (DD's) interface better, because I could move notes and I could move between notes more easily. LIke on the other one I'd have to click the back button, or the save button. But accidentally deleted a bunch of notes, but there's an undo button I guess."
- "I really wish I could add/split notes. On twinkle twinkle all the tones that there are two of them there was only one of them."
- phone crashes after export again.
Could you tell what's disabled?
...