Watching Themselves Play · Coding & portraiture appendix
A grounded coding scheme built from what they actually said, and portraits of individual musicians — drawn only from their own words.
Every tag, code, topic, and theme below was named and defined from words that actually appear in the transcripts — in-vivo (the musicians' own phrasing) where possible — each with verbatim examples carrying a timestamp and a name. Nothing was imposed in advance; categories with no real example were not invented.
Built bottom-up from 8 real-speech transcripts (Jennifer, Matt, Corad, Philippe, Yipeng, David, Khue, Nicole Collins). Automatic ASR, unverified. Every code below is derived from words that actually appear in the transcripts; categories with no clear real example were not invented. Codes are named in-vivo (participants' own phrasing) where possible.
Setting reminder (obeyed throughout): this is a research self-review — each musician watches themselves on live 360°/iPad multicam and thinks aloud. There is no teacher/student and no lesson. The second voice is the researcher, who runs the technology and prompts but never instructs or evaluates the playing. Speaker attribution below is INFERRED from content, not certain (single room mic, no diarization).
Confidence convention used while coding: lines with high agree + high v were treated as trustworthy; lines with v low, agree=0, or in the fabrication-flagged blocks were weighed with care or excluded (see HONESTY / LIMITS).
Small, close-to-the-words actions. Each is something a coder can point at in a single line.
T1. Names a body part they are watching.
T2. Requests / chooses a specific angle or view.
T3. Aesthetic delight / novelty reaction to seeing self.
T4. Names / complains about the delay (latency of the feed).
T5. Compares the feed to a mirror (home-practice reference).
T6. States a surprise / a disconfirmed belief about own body.
T7. Recalls a past external correction (mom / teacher / audience gaze).
T8. Declares a self-observation habit / where their eyes normally go.
T9. Rates self against an ideal ("how close to ideal?").
T10. Expresses discomfort / aversion at seeing self.
T11. Asserts sound over sight (modality hierarchy).
T12. Reports being distracted / can't help predicting own movement.
**T13. Notes the feed helps a specific goal (intonation / posture / technique).**
**T14. Says an angle shows nothing / declines it (a "no effect" report).**
T15. Checks appearance / grooming (hair, clothes) — a vanity glance.
**T16. Frames the observation as externalizing a felt technique (see what I usually only feel).**
T17. Notices sound is shaped by physical placement.
T18. (Researcher) prompts for noticing.
T19. (Researcher) frames setup / logistics / recording.
C1. Directed self-observation (tags T1, T2, T13, T16, T17) — Definition: the participant targets a named body part / technique and requests the angle that reveals it, treating the camera as an instrument for making felt technique visible.
C2. "I like this view" — angle preference as aesthetic + identity choice (tags T2, T3, T8) — Definition: choosing/enjoying an angle in a way tied to the player's discipline and self-image, not just utility.
C3. Latency friction (tags T4, T12) — Definition: the delay is named as breaking the loop — "disconnected," "overstimulating," or actively misread as the player's own fault.
C4. Mirror as the baseline frame (tag T5) — Definition: participants reason about the multicam by contrast with the home mirror — its one front-on view is the reference the extra angles are measured against.
C5. Belief revision through seeing (tags T6, T13) — Definition: the feed disconfirms (or confirms) a feared self-image; for skilled players it reassures more than it corrects.
C6. The internalized audience / authority (tag T7) — Definition: self-observation is filtered through a remembered corrector — mom, teacher, "how people view me," the stage.
C7. Self-evaluation against an ideal (tag T9) — Definition: rating current playing/body versus an imagined ideal performance, usually on a numeric scale prompted by the researcher.
C8. Aversion / self-consciousness at one's own image (tags T10, T15) — Definition: seeing oneself is emotionally charged; discomfort, grooming-checks, and a wish not to be seen.
C9. Sound-primary resistance to the visual (tag T11) — Definition: for ear-led musicians the visual is secondary or even a threat to the real (auditory) work.
C10. Overload / angle-count management (tags T12, T14) — Definition: negotiating how many simultaneous views are useful before they distract; "one at a time" vs the full grid.
C11. Researcher scaffolding (tags T18, T19) — Definition: the researcher's own register — logistics, consent, baseline/intervention framing, non-judgment, and noticing-prompts that structure the session.
| # | Topic | Recurs in | Anchor quote |
|---|---|---|---|
| TP1 | Camera angle choice & preference | all 8 | [12:23] Jennifer: "I want to start with that one." |
| TP2 | The delay / wireless latency | Jennifer, Matt, Corad, Philippe, Yipeng, David, (Nicole, Khue implied) | [05:00] Yipeng: "Oh, we have like a slower response rate." |
| TP3 | Posture / slouch / straight back | Jennifer, Matt, Philippe, Yipeng, Khue, Nicole, (David) | [25:19] Matt: "I'm trying to keep a straight back." |
| TP4 | Hands / fingers / slide / bow — discipline-specific technique | Jennifer (fingers, bow), Corad (thumb, fretboard), Philippe (stick symmetry), David (slide position), Khue (chord shifts) | [26:59] Corad: "keeping the thumb here is the biggest thing." |
| TP5 | Face / mouth / jaw / expression | Nicole (jaw, soft palate), David (cheeks, face), Yipeng (mouth/embouchure), Matt (facial expression), Corad (base face) | [35:44] Nicole: "raising the soft palate and trying to lower my tongue... to make more space for that high note." |
| TP6 | The mirror (home reference) | Jennifer, Matt, Nicole, Philippe, Khue | [56:x] Jennifer: "this is like even better than the mirror... you can have like different views." |
| TP7 | Discomfort / liking-vs-disliking watching oneself | Nicole, David, (Jennifer's vanity glance, Corad's "base face") | [41:00] researcher (David session): "40 to 60% of people that doesn't like to watch themselves." |
| TP8 | Sound vs. visual priority | Jennifer, Philippe, Yipeng | [58:57] Jennifer: "I'm focused more on the sound of my playing." |
| TP9 | Distraction / overstimulation / too many angles | Philippe, Nicole, David, Yipeng | [49:53] Nicole: "It is very distracting." |
| TP10 | Self-rating vs. ideal performance | Matt (7/10), Khue (5/10), Nicole, Philippe | [29:03] researcher (Philippe): "how close to your ideal performance to what you actually see?" |
| TP11 | Stage / audience perspective | Jennifer, Corad, Khue, Nicole | [39:33] Jennifer: "this is what people view me as." |
| TP12 | Setup / recording / wired-vs-wireless / consent (researcher-led) | all 8 | [57:04] Jennifer session: "So the moment you want to completely sync, it has to be wired." |
| TP13 | Where sound comes from physically (contact point / hand position) | Jennifer (contact point), Corad (pickup position) | [18:16] Corad: "where you play kind of changes the sound of it." |
TH1. "It shows me my fingers without having to look at my fingers" — the multicam externalizes proprioception. The strongest cross-participant pattern: musicians use the feed to see what they normally only feel, and to see it from angles their own eyes can never reach.
TH2. The delay both enables and sabotages self-observation. Low-stakes reflection survives the lag, but real-time correction dies in it — and one participant misattributes the lag to his own body.
TH3. Self-observation is filtered through an internalized audience. What they look for, and how they judge it, is inherited from mothers, teachers, and imagined stages. The multicam reactivates old corrections.
TH4. For skilled players the feedback reassures more than it corrects. Seeing disconfirms a feared self-image; the dominant affect is relief, not alarm.
TH5. Sound-primary musicians hold the visual at arm's length. Several rank hearing first and treat the image as secondary, distracting, or irrelevant to the "real" work — and skill level modulates whether the visual helps at all.
TH6. Watching yourself is emotionally loaded — from delight to dread. The same act produces "Ooh, love it" in one player and "blech / my worst feeling" in another; grooming-checks and self-deprecation ("my face!") sit in between. Vulnerability is a first-class finding.
TH7. Angle preference maps onto what each discipline cares about. Close-ups for hand technique (violin, bass, drums, trombone); front/mirror for singers (face + expression); side for posture; overhead for symmetry. Angle choice is a legible signal of the player's priorities — which is exactly what the researcher treats as the data.
TH8. Novelty of the unreachable view. A distinct pleasure/intrigue attaches specifically to angles the musician has never had access to (back, overhead, eye-of-the-instrument), separate from utility.
Attribution is inferred from content only (one room mic, no diarization). Treat as a tendency, not a certainty.
RESEARCHER tends to:
PARTICIPANT tends to:
Genuinely ambiguous lines (short backchannels — "Yeah," "Okay," "Interesting," "I see, I see") recur from both sides and were not attributed.
agree and voice-presence v are evidence, not proof. I built codes on high-agree/high-v lines and treated agree≤1 or low-v lines as weak.v, high m — e.g. Matt's "Everybody says don't," Nicole's "Cry Me a River," Khue's "How Far I'll Go") as reflection; they are performance.agree run-on blobs; I coded only clearly legible fragments (hand transitions, posture, 5/10 self-rating, "looking at the camera made me bend my head"). Much of that session is under-coded on purpose. Jennifer and Matt have long high-confidence stretches and are the best-evidenced.agree=2/4; several Philippe self-observation lines sit at agree=1/4). They illustrate but should not carry a theme alone.Each portrait has a segment-coding table (their meaningful real lines in time order, mapped to codebook ids) and a prose portrait of how that musician watched themselves — every claim anchored to a quote. They are deliberately different people: a sound-first violinist-pianist, a singer whose instrument is his own body, and a two-instrument player who read the feed's delay as his own slowness.
Source: asr_eval/reading/Jennifer.txt (automatic ASR, unverified; v=voice-presence, m=music, agree=# of 4 engines corroborating). Coded against asr_eval/coding/CODEBOOK_grounded.md.
Setting (obeyed): research self-review — Jennifer watches herself on live 360°/iPad multicam and thinks aloud. No teacher/student, no lesson. The second voice is the researcher (runs the gear, prompts, apologizes for the delay; never instructs or evaluates). Speaker attribution below is INFERRED from content only (single room mic, no diarization) — I mark the researcher's lines explicitly; unmarked rows are inferred to be Jennifer.
Confidence rule used: I coded lines with high agree + high v as trustworthy, weighed agree≤1/low-v lines with care, and excluded the fabrication-flagged block (see note at end). Jennifer is one of the best-evidenced sessions — long stretches sit at agree=3/4, v≈0.85.
| [mm:ss] | Her words (verbatim) | Code | Note |
|---|---|---|---|
| 10:58 | "I usually like move around, but I'm trying not to move around." | T8 | Names her own body habit before she has even seen the feed; the researcher [11:05] reassures "that's why that's 360 camera." |
| 13:07 | "I don't think I've ever seen myself in that view before." | T3 / TH8 | First look. The pleasure is specifically about the unreachable angle. |
| 13:09 | "Ooh, love it." | T3 | Immediate delight — the codebook's canonical delight token. |
| 13:56 | "Also, there's a little bit of a delay." (researcher, inferred) | T4 / C11 | Researcher names the latency early and owns it as a gear limit. |
| 18:39 | "This is so cool." | T3 | Sustained novelty affect as a fourth view is added. |
| 18:42 | "I'm afraid to move." | T3 → in-vivo "afraid to move" | The image is so absorbing she freezes — a small aversion-of-disruption inside delight. No clean code; flag as in-vivo. |
| 18:49 | "Oh, I'm seeing my finger position, whether if it's too flat or not." | T1 / C1 | Directed self-observation: names the body part and the exact fault she's checking. |
| 20:05 | "when I used to practice violin, because I wanted to see the finger placement, I always looked at the fingerboard like this, right?" | T8 / C1 | Establishes her lifelong gaze habit. |
| 20:15 | "my mom would comment that my eyes would look like it's too focused… so it didn't look good on stage." | T7 / C6 | The internalized corrector. Her whole finger-gaze suppression traces here. |
| 20:38 | "So I don't look at my fingers when I practice." | T8 / C6 | The inherited rule, now a habit. |
| 20:40 | "for me to look at that, it shows me my fingers without having to look at my fingers." | T16 / TH1 | The strongest cross-participant theme, in her own words: the feed externalizes what she trained herself not to look at directly. |
| 21:02 | "Well, I think it helps with like intonation." | T13 | Names a concrete goal the feed serves. |
| 21:16 | "The only thing that's a little difficult about this is it's delayed, so I'm seeing it a little later, but it's really interesting to see." | T4 / C3 | Latency friction, but held in tension with genuine interest. |
| 21:24 | "That's why I like practice in front of a mirror… but this one kind of shows the view from my eyes." | T5 / C4 | Mirror as baseline; the multicam gives an eye-line the mirror can't. |
| 25:58 | "This is like very similar to what I look at when I'm at home, like— oh, oh, that's scary." | C4 / T10 | Recognition of the home-mirror view carries a flash of aversion ("scary"). |
| 26:07 | "Can I look at my back? … I don't know, I've never looked at it." | T2 / TH8 | Reaches for the never-seen angle. |
| 26:13 | "Okay, my hair is in place. That's good to know." | T15 / C8 | The one vanity glance — quick grooming check, then back to work. |
| 28:09 | "I always thought that like my shoulder kind of like leaned forward when I practice… but it's good to know that I don't think I'm leaning forward." | T6 / C5 | Belief revision: a feared posture image disconfirmed; relief. |
| 39:11 | "I was looking all over the place, actually. This one for posture… This one for my finger placement… this is like the stage view. Like this is what people view me as." | T2 / T7 / TH7 | Maps each angle to a purpose; the stage/audience angle framed through others' gaze. |
| 40:24 | "it's like real-life feedback, but it's delayed, so it's a little overstimulating in a way." | C3 | Codebook anchor for latency friction — "overstimulating." |
| 40:46 | "I do play it slower in some spots than I normally do." (to see herself better) | T12 / C1 | The observation reshapes the playing — she slows down to look. |
| 43:00 | "the way you chose, right, helped a lot of what you're looking at." (researcher, inferred) | C11 / TH7 | Researcher states that angle choice itself is the data — non-judgmental framing. |
| 46:56 | "I like this view. I don't know why." | C2 | Angle preference as pure aesthetic/identity, unexplained. |
| 51:33 | "I just wanted to see my bow angle… with how angular it is from the bridge." | T1 / C1 / TP4 | Discipline-specific technique target (bow-to-bridge angle). |
| 52:20 | "I think I am not as slouching… as I thought." | T6 / C5 / TH4 | Reassurance over correction — feared self-image softened. |
| 52:42 | "But I'm not slouching as much as I thought, which is a good thing." | T6 / C5 / TH4 | Codebook anchor line; relief made explicit. |
| 52:50 | "this shows me that I do need to work on my bowing angles." | T13 / C5 | The one genuine correction she accepts — not feared, just a known task re-surfaced. |
| 53:07 | "Like I feel like I'm being reminded of it." | C5 | The feed reactivates a standing goal rather than revealing something new. |
| 53:31 | "I'm trying to not look at my left hand because right now with the delay, it's kind of getting me confused." | T12 / C3 / TH2 | Latency actively sabotages real-time technique-watching. |
| 54:55 | "I hope no one looks at me from that side though." | T10 / C6 | Aversion tied to the imagined audience gaze from a low/side angle. |
| 56:46 | "this is like even better than the mirror… because you can move it and you can have like different views." | C4 / TP6 | Multicam surpasses the home mirror's single front view. |
| 58:57 | "I always record sound and not really much of a video because I'm focused more on the sound of my playing." | T11 / C9 / TH5 | Modality hierarchy: ear first. Codebook anchor. |
| 59:19 | "I was always focused on the sound, and that's it." | T11 / TH5 | Reasserts the sound-primary identity. |
| 59:34 | "I don't think I really cared about what I look like… it's more for like the posture, more for the technique, than like what do I look like." | T11 / C9 | Separates function (posture/technique) from appearance; disowns vanity. |
| 60:43 | "this is very interesting because like this is kind of making me focus on what I look like." | T11 / C9 | Admits the visual pulls her toward appearance — a mild discomfort with that pull [60:49 "that could be also distracting"]. |
| 72:00 | "I've never looked at myself in the mirror when I'm practicing piano. It's impossible… Piano, mirror on the side." | C4 / in-vivo "dark-room piano" | Piano has no mirror tradition for her; the multicam view is unprecedented for that instrument. |
| 72:09 | "that's just the one thing I had to do growing up when I was practicing piano was like to be in a dark room with no light." | in-vivo "dark-room piano" / T8 | Striking: her piano training was anti-visual — no code fits, so I flag it in-vivo. |
| 72:23 | "looking at this, that's— this is the exact opposite of what I'm used to do." | in-vivo "dark-room piano" | The piano self-view inverts a lifelong habit of not-looking. |
(New in-vivo codes proposed only where nothing fit: "afraid to move" [18:42] — absorption-freeze; and "dark-room piano" [72:00–72:23] — an anti-visual practice tradition the mirror-baseline code (C4) doesn't capture.)
Jennifer arrives already watching. Before the feed is even up she confesses a body habit — "I usually like move around, but I'm trying not to move around" [10:58] — and the researcher has to reassure her the 360° rig means she needn't hold still [11:05]. That opening tells you who she is here: a player braced for how she looks, quick to self-monitor, and ready to be corrected before anyone corrects her.
What she reaches to see is almost entirely hands and posture, never her face. The first thing she names is "my finger position, whether if it's too flat or not" [18:49]; later it's "my bow angle… how angular it is from the bridge" [51:33], the contact point, the left hand. Her most revealing line is proprioceptive: "for me to look at that, it shows me my fingers without having to look at my fingers" [20:40]. That phrasing is not accidental. As a child she watched her own fingerboard until her mother told her "my eyes would look like it's too focused… so it didn't look good on stage" [20:15] — and "ever since then… I've been like very conscious of like not looking at my fingers" [20:33, 20:38]. The multicam resolves a decades-old bind: it lets her see the fingers she taught herself not to look at. Her self-observation runs entirely through her mother's stage-verdict and, behind it, the imagined audience — "this is the stage view. Like this is what people view me as" [39:33], and, of a low side angle, "I hope no one looks at me from that side though" [54:55].
Her reactions are mostly delight and relief, not aversion. First unreachable view: "I don't think I've ever seen myself in that view before" [13:07] → "Ooh, love it" [13:09] → "This is so cool" [18:39], so absorbed she reports "I'm afraid to move" [18:42]. Aversion surfaces only in flickers — the home-mirror view is "scary" for a beat before "No, I like it" [25:58–26:06], and one grooming glance, "my hair is in place. That's good to know" [26:13]. The dominant affect is reassurance: she'd long feared a slouch and a forward-leaning shoulder, and the feed disconfirms both — "I'm not slouching as much as I thought, which is a good thing" [52:42]; "it's good to know that I don't think I'm leaning forward" [28:09]. The one thing she lets it correct is not a fear but a standing task: "this shows me that I do need to work on my bowing angles… I feel like I'm being reminded of it" [52:50, 53:07].
The multicam let her see what a single mirror can't: "this is like even better than the mirror… because you can move it and you can have like different views" [56:46]. The delay cuts the other way. She likes the low-detail reflection but names the lag repeatedly — "it's delayed, so I'm seeing it a little later" [21:16], "it's a little overstimulating" [40:24] — and it defeats real-time technique-watching: "I'm trying to not look at my left hand because right now with the delay, it's kind of getting me confused" [53:31]. She even adapts her playing to the feed, "I do play it slower in some spots" [40:46].
Underneath, she filters all of this through a sound-first identity: "I always record sound and not really much of a video because I'm focused more on the sound of my playing" [58:57], "I was always focused on the sound, and that's it" [59:19]. The visual is a guest — welcome, illuminating, faintly suspect: "this is kind of making me focus on what I look like… that could be also distracting" [60:43, 60:49]. The session's quiet surprise comes at the piano, where she has no mirror tradition at all: "one thing I had to do growing up… was like to be in a dark room with no light" [72:09], so seeing herself play piano is "the exact opposite of what I'm used to do" [72:23]. For the violinist the camera restores a forbidden gaze; for the pianist it breaks a lifetime of playing blind.
Honesty note. Jennifer's key lines sit at agree=3/4, v≈0.85 — strong. The stretch from ~48:00–56:00 has very low v (0.05) with the substance still corroborated at agree=3/4, so I trusted content but not the exact wording there. Much of 02:00–11:00 and the "© transcript Emily Beynon" / "Thank you" / "I love it ×8" blocks are fabrication artifacts (transcript lines 636–1036) and were excluded; notably a real "Ooh, love it" [13:09] sits far from the hallucinated "I love it" loop, and the coded delight rests on the real one.
Coded 2026-08-03. Grounded codebook v1. All claims are anchored to timestamped lines; speaker attribution is inferred from content, not verified.
Research self-review: Matt watches himself on live 360°/iPad multicam and thinks aloud. There is no teacher, no student, no lesson. The second voice is the researcher, who runs the rig and prompts but never instructs or evaluates. Source: asr_eval/reading/Matt.txt (camera 360-Camera-2, automatic ASR, unverified). Speaker attribution is INFERRED from content — single room mic, no diarization. Codebook: asr_eval/coding/CODEBOOK_grounded.md. Sung-lyric lines (low v, high m) are performance, not reflection, and are not coded as reflection per codebook LIMITS. Fabrication-flagged lines (Matt.txt 512–880: REP×N loops, hallucinated "Thank you"/"Bye") are excluded.
Time-ordered coding of Matt's meaningful real reflective lines. All rows below are inferred to be Matt (participant) from first-person body/voice content unless noted. Confidence = the transcript's own agree/v.
| [mm:ss] | His words (verbatim) | Tag/Code (codebook id) | Note |
|---|---|---|---|
| 03:38 | "it's because the whole body's used for voice, so I just like to make sure it's at least a little loose." | in-vivo: BODY-IS-INSTRUMENT (no codebook code fits; nearest T16) | Stretching before warm-up; for a singer the body is the instrument. agree 3/4, v .86. |
| 03:45 | "Yeah, all instrumentalists should do that." | in-vivo: BODY-IS-INSTRUMENT | Positions himself alongside instrumentalists but with body-as-instrument. 3/4. |
| 06:21 | "there just seems to be a little something in the higher range that just feels disconnected today for whatever reason." | in-vivo: VOICE-FEELS-DISCONNECTED (felt/auditory vocal state; no fit) | Reports the instrument by feel, before watching. 3/4, v .81. |
| 06:32 | "it doesn't feel like it's vibrating exactly as I ideally want it to… and sometimes that's just the way it is that day and you kind of just work through it." | in-vivo: VOICE-FEELS-DISCONNECTED | Proprioceptive/auditory, not visual. Nearest T11/C9 (sound over sight). 3/4. |
| 06:58 | "I haven't sung too, too much this summer, so it's coming out of… just haven't used the instrument too much." | in-vivo: BODY-IS-INSTRUMENT | "the instrument" = his voice. 3/4. |
| 14:54 | "one thing I'm looking out for, I've been told sometimes… especially for like acting things out, sometimes I like do this, and apparently that's like not good." | T7 (recalls past external correction) → C6 internalized authority | Inferred Matt. 3/4, v .76. |
| 15:07 | "like Jeannie told me one time it was like I lead with the head or something like that." | T7 / C6 | Named corrector (Jeannie). 3/4. |
| 15:13 | "So I'm trying to be conscious of that… from every angle so that I don't slouch or do this or do that." | C1 directed self-observation / TP3 posture | 3/4, v .80. |
| 15:19 | "that's why I wanted to have it front and back, just to have like the full idea of it." | T2 requests angle / C1 | Chooses angles to serve posture goal. 3/4. |
| 15:28 → 15:31 | "I think I did a good job." … "it was good to have the views to judge that a little bit better." | C5 belief revision (reassurance) / T13 | Feed confirms, doesn't alarm. 3/4. |
| 15:44 | "from this angle, from this… camera number 2, it's a little bit harder to tell." | T2 angle utility (nuance) | Side view less legible than front. 3/4. |
| 15:51 | "I liked having the frontal view just to see myself… but it's good to have the side view." | C2 angle preference / T2 | 3/4, v .82. |
| 17:39 | "Oh, I love it." | T3 delight / novelty | Inferred Matt, reacting to the back view. 3/4. |
| 21:38 | "I have the main things I'm trying to look at, I can see." | C1 directed self-observation | 3/4, v .82. |
| 22:11 | "Oh shit, I forgot the words." | performance (not coded as reflection) | Mid-song break; context only. 3/4, m .73. |
| 23:27 | "I always screw up that part." | performance-adjacent self-observation of a habitual error | Weak: agree 2/4, high m. Weigh with care. |
| 25:11 → 25:15 | "It's hard to tell where my body naturally kind of is." … "right now, I don't know exactly where my back is and where it just kind of floats." | in-vivo: BODY-FLOATS + T16 (externalize the felt) | Black long shirt removes his body's visual outline; proprioception alone is uncertain. 3/4, v .83. |
| 25:19 | "sometimes I feel like I have a straightened-out back and I don't." | T6 disconfirmed belief / C5 | Core belief-revision moment. 3/4. |
| 25:31 → 25:36 | "if I was just looking at… my back, I wouldn't be able to tell that. But since I see these other 2, I can see that it's a little extended." | TH1 externalize proprioception / C1 | Multicam does what one view can't. 3/4. |
| 25:57 | "it's one thing for someone to tell you to straighten out, but it's another thing to like know what that feels like and to see it." | T13 (codebook anchor) | Bridges feel ↔ see. 3/4, v .86. |
| 26:05 | "when you're just looking in a mirror, you only see like the front view." | T5 mirror comparison / C4 (anchor) | 3/4, v .88. |
| 26:09 | "now I can see what it looks like from everything and adjust it exactly where it needs to be… I feel like I would get used to the… way that that stance feels." | TH1 / C4 | Wants to internalize the corrected stance. 3/4. |
| 26:38 → 26:42 | "it is cool to see all 3. I do like that." … "Yeah, I do." | C2 / C10 counter-case (no overload) | Embraces 3 simultaneous angles — unlike the overload theme. 3/4. |
| 27:08 → 27:11 | "once I get my center, I could be a little bit more free with it… check in with my body… the posture that allows for the easiest singing possible." | in-vivo: CENTER-THEN-FREE / T13 | Posture as the enabler of free singing. 3/4. |
| 30:23 | "in singing you're never going to be completely straight all the time." | TP3 posture (nuance) | Rejects rigid ideal. 3/4. |
| 32:19 | "I did the same thing, and this was kind of confirming that it looks okay." | C5 / TH4 reassurance | Second take confirms the first. 3/4. |
| 33:04 | "7 out of 10." | T9 / C7 (anchor) | Self-rating vs. ideal, researcher-prompted. 3/4, v .91. |
| 33:14 → 33:32 | "my primary time of… voice lessons was undergrad, and I used to— I still do it sometimes— I had a lot of tension right here… something I was always conscientious of, just trying to be more loose." | T7 / C6 past correctors / TP5 neck | Undergrad teachers still in the room. 3/4. |
| 34:05 → 34:13 | "I'm picturing like a violin player overhead… for this, I think this is— if I had 3 cameras set up, I think I would set them up like this." | T14 declines an angle / TH7 angle↔discipline | Overhead is for instrumentalists, not him. 3/4. |
| 34:29 → 34:34 | "facial expression… that's the one thing you do always get out of the mirror because it's usually right in front of you. So that's not something I was looking at too much with this." | TP5 face / T5 / C4 | Mirror already covers the face → he spent multicam on posture/back. 3/4. |
| 34:38 | "I didn't want to like be looking at my mouth moving at different times than I actually was because I would get even more confused than I already am." | T4 delay / C3 latency friction / TH2 | The singer's delay answer: he avoided the mouth to escape the lag. 3/4, v .76. |
A singer whose instrument he cannot hold at arm's length. Where a violinist can point the camera at the fretboard, Matt's instrument is the body doing the watching. He says it plainly during warm-up: "the whole body's used for voice, so I just like to make sure it's at least a little loose" [03:38], and when he apologizes for a rough high range it is the instrument, not the throat, that he blames — "I haven't sung too, too much this summer… just haven't used the instrument too much" [06:58]. This is the frame for everything he looks for. He cannot aim the multicam at a thing; he can only aim it back at himself.
What he looked for: posture, not face. Almost every reflective line is about the line of his body — back, neck, "center." That choice is deliberate and, tellingly, defined against the mirror. Facial expression, he explains, "is the one thing you do always get out of the mirror because it's usually right in front of you. So that's not something I was looking at too much with this" [34:29–34:34]. The mirror already gives him his face; the multicam is worth its trouble only for the views a mirror can't reach. So he spends it on posture, front and back and side, "just to have like the full idea of it" [15:19].
What the multicam revealed that a mirror can't — and the black-shirt problem. The most vivid moment is a failure of proprioception made visible. In a long dark shirt he loses his own outline: "it's hard to tell where my body naturally kind of is… I don't know exactly where my back is and where it just kind of floats" [25:11–25:15]. His felt sense is unreliable — "sometimes I feel like I have a straightened-out back and I don't" [25:19] — and here the extra cameras do their real work: "if I was just looking at my back, I wouldn't be able to tell that. But since I see these other 2, I can see that it's a little extended" [25:31–25:36]. He names the whole point of the study better than the codebook does: "it's one thing for someone to tell you to straighten out, but it's another thing to like know what that feels like and to see it" [25:57], "when you're just looking in a mirror, you only see like the front view" [26:05]. And crucially he wants to keep it — to "get used to the way that that stance feels" [26:09]. Sight is a scaffold for a new proprioception.
How it felt: confident, not exposed. Matt is one of the reassured, not the dreading. He watches three angles at once and, against the overload theme, likes it — "it is cool to see all 3. I do like that" [26:38]. He grades himself calmly at "7 out of 10" [33:04] and reads the second take as "confirming that it looks okay" [32:19]. No "blech," no vanity flinch; the affect is a working musician checking his instrument.
Did the delay matter to a singer? Yes — and he routed around it. For instrumentalists the lag sabotages hand-watching. Matt's target could have been the mouth, and precisely there he refused: "I didn't want to like be looking at my mouth moving at different times than I actually was because I would get even more confused than I already am" [34:38]. He kept to posture — a slow, near-static target the delay can't spoil — and left the fast-moving mouth to the mirror. The lag shaped his method rather than defeating it.
An internalized studio. His eyes go where past teachers pointed them. Jeannie's "I lead with the head" [15:07] and undergrad voice lessons where "I had a lot of tension right here" [33:14] are the corrections he's now re-checking from every angle. The end state he wants is not stiffness but freedom on a stable base — "once I get my center, I could be a little bit more free with it… the posture that allows for the easiest singing possible" [27:08–27:11] — remembering that "in singing you're never going to be completely straight all the time" [30:23].
ASR caveat: attribution is inferred from content; the long reflective stretches quoted (25:11–26:42, 33:04–34:38) run at agree 3/4 and v ≈ .8, the strongest evidence in the file. Sung lines and the fabrication block were excluded.
Multicam self-review, camera 360-Camera-2. Automatic ASR, unverified. Speaker attribution is INFERRED from content only (single room mic, no diarization) — treat "(researcher)" as a content-based guess, not certainty. Reliability leans on lines with high cross-engine agree (of 4) and high voice-presence v; low-v/agree≤1 lines are used only as color and flagged. The instrument David calls the đàn tranh is mangled by the ASR as "chanbo," "cheongwon," "chanbo inside," and the finger-picks as "these" — read accordingly. The proper noun "Chongwon" (~33:01) is ASR-uncertain and may be a person/ensemble name or a mishearing.
Time-ordered coding of David's meaningful real lines (participant unless marked). Codebook ids: T# tags, C# codes, TP# topics, TH# themes. In-vivo codes are labeled IV and flagged as such.
| [mm:ss] | His words (verbatim) | Tag/code | Note |
|---|---|---|---|
| 01:00 | "So you'll see me breathe a lot, because there's no other trombonist with me." | TP4 | Baseline self-description; frames what the camera will show before it is on. agree 3/4. |
| 01:37 | "So I'm playing the bass trombone, uh, which I'm— is not technically the trombone I'm playing, which is the tenor trombone." | TP4 | Instrument identity; sight-reading a bass part on a tenor. |
| 05:53 | "it's more like I like trying to be like, oh, it's a staccato, so I'm gonna play with my tongue instead." | TP4 | Technique talk (articulation). |
| 06:07 | "staccato accent, I still never get it really." → 06:13 "I'm like loud and short." | TP4 / IV self-critique | Honest admission of a lasting weakness (no numeric rating, so not clean C7). |
| 19:33 | "that's me pretending to sight-read through it." | IV "pretending to sight-read" | Watching the playback he narrates his own bluffing; no exact codebook fit. |
| 20:55 | "It's been 1 and a half years, but I sight-read this bass part because I don't play bass, I play trombone too." | TP4 | Context for the miscounting/dragging talk that follows. |
| 21:25 | "usually I would find like a composition online, or... a YouTube video of like a playback." → 21:35 "there's like a score, and then it'll just like track it. So I like play along with it." | TP4 | Habitual practice = play along with an external track (foreshadows reliance on an external tempo). |
| 23:54 | (researcher, inferred) "Since you're working on that, shall we just turn one of the cameras on just to see if you actually see something?" | T18 / C11 | Intervention start — the live feed comes on. |
| 26:22 | "if I want to like watch my slide position Then it'd be really hard." | C1 / T2 | Directed self-observation: names the slide as the thing to watch; notes it's awkward to see. |
| 27:55 | "Wow, I forgot my muscle memory all of a sudden." | T12 / C3 | Watching himself breaks automaticity — the loop begins to interfere. agree 3/4, v 0.84. |
| 29:22 | "And it's like, uh, delayed a bit, so I'm like, ah." | T4 / C3 | Names the latency. (Also at 28:05.) |
| 32:37 | "I saw— well, I look up and down a bit." | T8 | Reports where his eyes go; adds "That's when I usually have bad notes" (32:24, merged line). |
| 32:55 | "I know that's a delay." | T4 | He explicitly recognizes it is latency. agree 3/4, v 0.88. |
| 32:56 | "I know, I'm like, I'm too slow, I need to be faster." | C3 / TH2 | Key moment (codebook-flagged): reads the feed's lag as his own slowness. High confidence: agree 3/4, v 0.90. |
| 33:01 | "Yeah, no, Chongwon's well known for dragging." → 33:03 "I like when people say you're dragging, I'm like, I don't know." | C6 / TH3 | The lag lands on a pre-existing reputation for dragging — internalized audience. ("Chongwon" ASR-uncertain.) |
| 33:07 | "And it's like, oh yeah, I am slow." → 33:17 "that's what I'm telling myself." | C3 / TH2 | The misattribution consolidates into a verdict on himself. |
| 33:25 | "Maybe because I'm self— I need to look at myself a little more, not the trombone." → 33:29 "I look at the trombone." | T8 / C1 | On trombone his gaze normally goes to the instrument, not his body; the feed prompts a shift. |
| 34:18 | "I got nervous because I'm like, oh, it's delayed, slow, and too slow, I gotta be faster." | C3 / TH2 / T10 | Clearest statement of the loop: the delay produced nervousness, not just a note. agree 3/4. |
| 34:31 | (researcher, inferred) "No, you're already super fast." | C11 | A human, not the feed, breaks the misread and reassures him. |
| 39:23 | "I will say it does not help with memorization." | T14 | Honest "no-effect" report for one goal. |
| 39:50 | "trying to tie, um, the pitch and the mouth position." | T13 / C1 | Where the feed does help: linking pitch to embouchure. |
| 40:28 | "mostly the top-down version." → 40:30 "Because I'm like, my face!" → 40:42 "top-down is okay... it's a little less self-conscious." | T10 / C8 / TP7 | Chooses the overhead angle to avoid seeing his own face; delivered as comic self-deprecation. |
| 41:05 | "I don't hate watching myself because I have to watch myself anyway." | C8 (counter-case) / TP7 | David is camera-comfortable (speeches, YouTube, teaching) — vulnerability shows as comedy, not dread. |
| 41:48 | "I sometimes don't hold it completely right." → 41:57 "it does make it faster when you just wrist... instead of like with my arm." → 42:05 "I see that a little bit, like, ah, do that, stop doing this." | C1 / T16 / TH1 | Feed working as intended: externalizes a felt technique (grip, wrist-vs-arm motion). |
| 42:35 | "I would tell my friends, yeah, look, We are dragging." → 42:44 "I am dragging." → 42:45 "I am so sorry, y'all." | C6 | Dragging theme returns, now owned and performed to an imagined crowd. |
| 47:06 | "No, I'm self-conscious now about puffy cheeks." → 47:36 "you can see my cheek puffed out." | T10 / C8 / T1 / TP5 | Camera surfaces a face-habit; names the body part; charged affect. |
| 49:07 | "in the 6th grade, teacher's like, no, don't [puff] cheeks, stop it." → 49:34 "society doesn't like you puffing out your cheeks." | C6 / TH3 | Puffy-cheek self-consciousness traces to an old teacher's correction + a societal gaze. |
| 51:21 | "It's very confusing to see 3 at the same time." → 51:07 "Or you rather see one, right?" | C10 | Angle-count overload (codebook anchor for C10). |
| 56:35 | "I just know it's there in position." → 56:17 "I don't look at it anymore." | T16 / TH1 | Trombone slide is felt, not seen — proprioception the camera could externalize. |
| 69:47 | "This, um, light instrument, the resonance is really nice." → 69:57 "It's like it's light and it vibrates a lot." | T11 / TP8 / IV resonance-delight | Đàn tranh: affect flips to warm; delight is in the sound of the instrument, not in seeing himself. |
| 80:02 | (researcher, inferred 79:32 "Your right hand definitely a lot more fine-tuned control, right?") David: "my left hand still ever always needs work, but it's fine." | TP4 / IV left-hand-underdog | A cross-instrument bodily fact named aloud. |
| 80:38 | "it's good because I totally can see, uh, me land perfectly more." | T13 / C1 / TH1 | On đàn tranh the feed helps — he watches his hand land. |
| 80:43 | "I don't use my left hand at all because my trombone is like gripped there." → 81:00 "when you see... me playing piano, it's like my right hand's fine." | IV left-hand-underdog (cross-instrument) | Multicam makes legible one truth spanning trombone, piano, zither; no single codebook cell fits. |
| 86:14 | "oftentimes my fingers just feel like going Throw this... it just pulls on the string like this and it flings off." | T16 / TP4 | Watching reveals his plucking flings the finger-picks off; "lost 2 of these in the woods" (86:23). |
David arrives at his own image the way a showman arrives on stage: talking. He narrates to a crowd that isn't in the room — "I would tell my friends, yeah, look, We are dragging" (42:35), "I am so sorry, y'all" (42:45), "I'm gonna send this to my friends" (58:12), "upload this for YouTube" (58:16). When the codebook's cited fact surfaces — that "40 to 60% of people... don't like to watch themselves" — David places himself outside it: "I don't hate watching myself because I have to watch myself anyway" (41:05). So his vulnerability rarely reads as dread. It reads as comedy: "Because I'm like, my face!" (40:30), "look at me play drama" (40:36). The laugh is the armor.
What he reached to see differed sharply by instrument. On trombone he does not, ordinarily, watch himself at all — "I look at the trombone" (33:29); the seven slide positions are felt, not seen: "I just know it's there in position" (56:35), "I don't look at it anymore" (56:17). The camera's novelty is exactly that it threatens to turn felt into seen, and he flinches: "if I want to like watch my slide position Then it'd be really hard" (26:22), then, moments after the feed comes on, "Wow, I forgot my muscle memory all of a sudden" (27:55). On đàn tranh the same act is warmer and more welcome. He improvises, delights in the instrument itself — "this light instrument, the resonance is really nice... it's light and it vibrates a lot" (69:47, 69:57) — and lets the feed help: "I totally can see, uh, me land perfectly more" (80:38). Across both instruments the multicam surfaces one bodily truth he says out loud: his left hand is always the underdog — "my left hand still ever always needs work" on the zither (80:02), "I don't use my left hand at all because my trombone is like gripped there" (80:43), "my right hand's fine" at the piano (81:00). One session made legible a fact about his whole body, not one instrument.
The delay is the heart of it, and it is honestly, unambiguously real — these are among the best-evidenced lines in the cohort (v ≈ 0.9, agree 3/4). The striking thing is that awareness did not protect him. He names the latency plainly — "I know that's a delay" (32:55) — and in the very same breath absorbs it as a verdict on his body: "I'm too slow, I need to be faster" (32:56), "oh yeah, I am slow" (33:07), "that's what I'm telling myself" (33:17), and finally, "I got nervous because... it's delayed, slow, and too slow, I gotta be faster" (34:18). This is the feedback loop misleading a self-aware performer. It stuck because it did not arrive on blank ground — it landed on an existing wound: "Chongwon's well known for dragging... I like when people say you're dragging, I'm like, I don't know" (33:01–33:03; the name is ASR-uncertain). The lag spoke in the voice of an old critic, and David, primed to believe he drags, believed the machine. That the feed can impersonate a real insecurity — not invent a new one — is the precise way it can distort. And note who dissolved the illusion: not the technology, but the researcher in the room — "No, you're already super fast" (34:31). The delay manufactured the doubt; a person had to unmake it.
Where the feed corrected cleanly, it did so on technique he normally only feels: "I sometimes don't hold it completely right" (41:48), "it does make it faster when you just wrist... instead of like with my arm" (41:57), "I see that a little bit, like, ah, do that, stop doing this" (42:05). And it reactivated a second internalized authority — an old teacher who policed his face: "self-conscious now about puffy cheeks" (47:06), traced straight to "6th grade, teacher's like, no, don't [puff] cheeks, stop it" (49:07) and a diffuse "society doesn't like you puffing out your cheeks" (49:34). Twice, then, the mirror showed him what a corrector once flagged — the tempo and the cheeks — which is why he retreats to the top-down view that hides the face (40:42) and calls three angles at once "very confusing" (51:21). David watched himself as a comedian watches a blooper reel, but under the jokes the feed reached two tender places, and its truest use was less to fix the notes than to hold a mirror to his hands and his self-image — most trustworthy when a human, not the delay, was there to say what was real.
asr_eval/: codebook coding/CODEBOOK_grounded.md · portraits
coding/portraits/ · clean reading transcripts reading/ (backbone AssemblyAI, cross-engine
agreement) · reliability evidence in the ASR appendix. Built by
clean_reading.py + LLM coding + build_coding_page.py. Toward the coding / portraiture
lane of the dissertation (separate from the ASR-reliability paper).