COHESIVE GOTHIC PIANO AND HARMONICA FUSION Composition study and bounded experiment plan 2026-11-07 05:03:29 UTC RESULT Use one intimate performance as the center: a youthful-sounding adult male voice, restrained piano and sparse harmonica in a small, imagined late-night urban room. Gothic melancholy shapes the harmony and emotional restraint; a blues-like looseness shapes the vocal and harmonica phrasing; a restrained 1980s-inspired production treatment supplies atmosphere. These influences should serve the same scene rather than appear as separate genre demonstrations. The current experiment target is FOUR MINUTES, within the latest 3.5–6-minute range. The earlier seven-minute target was superseded. Four minutes has demonstrated local generation evidence; longer performance remains untested here. This target applies to this experiment and does not change separate recurring-song requirements. This document is a composition plan and research result, not a rendered song or listening-approved master. Exact instrumental realization, emotional effect, vocal quality and lyric fidelity have not been established. “Masterpiece” remains an artistic aspiration, not a quality claim. 1. WHAT MAKES THE COMBINATION ONE SONG Keep three anchors stable: - One small-room perspective. The voice, piano and harmonica should sound as though they belong in the same space. A distant city texture is a background, not a second scene competing with the performer. - One recurring musical identity. Use a short piano figure that can be recognized when the singer or harmonica answers it. Change its timing, final note or voicing sparingly rather than introducing a new motif for every style label. - One emotional ceiling. Remain subdued, tired, melancholy and observant. Do not turn the final refrain into triumph, a major-key release, an arena chorus or a louder climax. Assign each influence a job. Piano supplies harmony, pulse and the principal motif. Voice supplies the human narrative. Harmonica supplies short replies and occasional wordless continuation. Effects and room texture supply time and place. If an added part duplicates one of those jobs without improving the song, remove it. This is an authored design method, not a scientific theorem of cohesion. Ableton's arrangement guidance describes subtraction as one way to turn available material into a time-based structure, and its discussion of unique events recommends restraint rather than constant novelty. Those are practical production strategies, not evidence that this specific arrangement will succeed. [1, 2] 2. TRANSLATE THE DESCRIPTORS INTO COMPATIBLE CONTROLS Dark/gothic/alternative: use minor-centered harmony, unresolved upper notes, reserved pacing and sparse space. The result need not imitate any artist or assume that minor harmony causes the same feeling in every listener. Low and not rising much: ask for a comfortable low vocal tessitura and narrow melodic movement, mostly neighboring notes. In an editable score, begin with roughly a fourth or fifth of range and adjust to the actual singer. This is a compositional limit, not a claim of literal vocal-fold inversion. Low chord inversions and restrained vocal register are different controls and should be described separately. Raspy/scratchy/smoky: use a modest rough edge around a supported, intelligible center. Do not pile harsh distortion, breath noise, whispering and heavy slurring onto every word. A small amount of texture has more contrast when some syllables remain clean. Drunken/off-beat/slurred: interpret these as a stylized, slightly late, unsteady-seeming delivery with softened phrase endings. Preserve the important consonants and return to the musical phrase boundary. Do not prescribe intoxication, random timing errors or a loss of every word's identity. Whispers/suppression: create intimacy with close perspective, soft delivery and brief breath-like inflections. Literal whispering need not dominate the recording. If a human performs the piece, do not force roughness or manufacture hoarseness; the desired sound can be approached with appropriate performance and controlled processing instead. NIDCD advises avoiding overuse, singing on a hoarse or tired voice, and vocal extremes including whispering. [3] Piano solo plus harmonica: make piano the continuous center and give it genuinely unaccompanied passages. Let harmonica enter only in selected gaps. This reconciles a piano-led solo atmosphere with the requested second instrument instead of treating them as incompatible instructions. Small audience and urban environment: use a faint, nonverbal room presence and distant urban ambience only if suitable cleared or synthetic material is available. Avoid intelligible private conversation, distracting applause or the implication that a synthetic scene is a real concert recording. No crescendos, no uplifting turn: hold the overall intensity stable. Small natural note dynamics can remain, but there should be no sustained build in volume, density, register or celebratory harmony. Development can occur through silence, timing, voicing and which part answers. 3. PROPOSED MUSICAL GRAMMAR Starting tempo: 60 quarter-note BPM, 4/4. This is a proposed slow pulse, not a measured tempo. Piano can imply it without a prominent drum track. Tonal center: D minor. One possible harmonic family is Dm(add9), B-flatmaj7/D, Gm(add9), and a return to Dm(add9). These choices keep related tones available without requiring a grand dominant-to-tonic resolution. The actual voicings must be checked for register and clarity. Piano: soft low-middle register, spaced bass notes rather than thick clustered low chords, and a three-note upper figure. Leave enough space for its decay. Do not equate low register with maximum bass energy or let sustained pedal blur every harmonic change. Voice: adult male, youthful timbre, low comfortable register, understated rasp, narrow contour, slightly behind the pulse. Use short comprehensible clauses and an understated repeated phrase. Avoid belting, high melodic leaps, dramatic vibrato and ornamental runs on the first statement of important information. Harmonica: restrained breath-shaped answers, often only a few notes, in a register that does not fight the words. It can echo the piano figure with a changed ending or hold a questioning note after the singer stops. Choose the actual instrument/key and playable notes before claiming exact score control. Production: keep the vocal close and relatively clear. Use a dark, low-level echo or short room treatment mainly in phrase gaps. A subtle chorused or softened piano tail can suggest an older electronic production color without adding a synth stack. These are production proposals, not claims about a definitive 1980s sound. Do not use heavy reverb as a substitute for sadness. Match effects to the phrase and check consonant clarity, low-frequency buildup and mono compatibility. Exact effect parameters require listening and cannot be guaranteed by a text caption. 4. FOUR-MINUTE FORM WITHOUT A BUILD The timeline below totals exactly 240 seconds and 60 bars at the proposed 60 BPM in 4/4. The arithmetic was checked. No corresponding audio has been generated for this study. 0:00–0:24, 6 bars: Piano alone establishes the recurring figure. Very faint room presence is optional. No cinematic swell. 0:24–1:08, 11 bars: First verse. Voice stays close and low. Piano answers some phrases but leaves others open. The slightly unequal phrase grouping is deliberate, not a timing error. 1:08–1:32, 6 bars: Restrained refrain. Keep comparable loudness and register. A small harmonic or rhythmic recontextualization supplies the difference. 1:32–2:00, 7 bars: Piano and a brief harmonica exchange. Let the last response arrive later, rather than increasing intensity. 2:00–2:44, 11 bars: Second verse. Reuse the central figure with one changed ending or bass voicing. The narrative adds a concrete observation, not a more theatrical emotional declaration. 2:44–3:12, 7 bars: Refrain variation. Remove a familiar answering note or shorten an echo. A missing gesture can make the return feel changed without a crescendo. 3:12–3:36, 6 bars: Piano alone. Keep the identity of the motif while allowing more space. Do not fill the break with unrelated virtuosity. 3:36–4:00, 6 bars: Quiet coda. A final low phrase may return; harmonica can leave a short unresolved answer. End with piano decay, not a surprise loud ending, crowd cheer or hopeful lift. Neural generation may not follow these boundaries. If precise form is essential, use an editable, score-led or stem-based assembly and verify transitions. Do not present a caption as proof of a 60-bar arrangement. 5. COMPACT GENERATION BRIEF Original four-minute gothic piano-noir song with restrained blues-harmonica replies and subtle 1980s-inspired room and echo treatment. A youthful-sounding adult male low voice, soft smoky rasp, narrow melodic range, slightly behind-the-beat delivery with gently blurred endings but intelligible key words. Intimate, weary, subdued and melancholy throughout. Soft low-middle piano carries one recurring three-note figure and the slow pulse; sparse harmonica enters only between selected vocal phrases. Small quiet urban room, faint nonverbal audience presence, distant city atmosphere. Include piano-alone passages. Closely related minor-centered harmony, restrained dark echoes, no sustained build, no crescendo, no belting, no high-register lift, no triumphant chorus, no dramatic solo or arena crowd. Develop through spacing, voicing, withheld replies and silence. Natural quiet ending. No artist imitation. This is a proposed input. It is not a transcript or a claim that every requested property will be produced. Lyrics and their underlying story still need to be written and reviewed separately. 6. FEASIBILITY AND HONEST DELIVERY Three limits must remain separate: - Upstream documentation: ACE-Step's interface describes duration settings up to 600 seconds, while also explaining that permitted duration is adjusted to the initialized hardware/configuration. That advertised range is not a local guarantee. [4] - Current local wrapper snapshot: the inspected runner accepts up to 360 seconds. It was not changed by this research. - Actual completed evidence: inspected generation records demonstrate 240-second output, with recent four-minute generation times around eleven minutes. They do not demonstrate a five-, six- or seven-minute successful render or perceptual quality. Four minutes fits the latest experiment range and the demonstrated duration. A shorter excerpt could be used first as a clearly labeled test, but it would not by itself complete the requested full-duration experiment. Do not pad a short render with repeated audio or silence and call it a newly composed longer song. Before delivery, verify actual duration, finite/non-silent audio, file formats, and preservation of original masters. Then separately check words, phrase transitions, instrumental roles, unwanted builds, vocal register and the quiet ending. ASR can flag uncertainty but cannot certify these musical properties. If listening is unavailable, state that limit and describe the output as technically checked rather than release-approved. 7. APPLY THE LEARNING TO OTHER COMPATIBLE FUSIONS Do not force this dark palette onto every future song. The reusable method is to choose a shared purpose, one rhythmic basis, a small harmonic vocabulary and clear roles before borrowing techniques. Possible later experiments, all untested: - Muted jazz harmony with coldwave restraint: keep a limited motif and steady sparse pulse; use richer voicings for color rather than constant harmonic novelty. - Quiet blues phrasing with minimal electronic texture: preserve the voice/instrument dialogue while letting a subtle electronic part support the same pulse. Avoid making the electronic section a separate dance track. - Gothic piano with restrained dub spacing: use selective delayed replies and bass continuity, preserving the lyric and the intended emotional ceiling. - Folk-like narrative with smoky alternative textures: keep clear storytelling and use tone, room and articulation to supply contrast rather than copying identifiable melodies or voices. For each experiment, keep two anchors stable and vary one principal influence first. Compare anonymous, level-matched excerpts when possible. Ask whether the blend feels coherent, whether its parts have distinct useful roles, whether the words remain understandable, and whether the emotional direction fits. Record disagreement and exceptions. These tests can guide taste; they do not establish universal compatibility or guaranteed originality. RESEARCH SOURCES [1] Dennis DeSantis / Ableton, Arranging as a Subtractive Process. Practical arrangement method; not an empirical guarantee. https://makingmusic.ableton.com/arranging-as-a-subtractive-process [2] Dennis DeSantis / Ableton, Unique Events. Practical guidance on sparing variations within repeated material. https://makingmusic.ableton.com/unique-events [3] National Institute on Deafness and Other Communication Disorders, Taking Care of Your Voice. Used only for the distinction between an artistic sound request and safe human vocal use. https://www.nidcd.nih.gov/health/taking-care-your-voice [4] ACE-Step 1.5, Gradio Demo User Guide. Duration settings and hardware-dependent initialization limits checked November 7, 2026. https://ace-step.github.io/ACE-Step-1.5/en/GRADIO_GUIDE Evidence status: current local runner and completed-output records were inspected; timeline arithmetic was checked. No model extension, new software installation, source-media upload, new audio render or listening experiment was performed in this research pass. The report contains only public-safe creative and technical findings.