Dream Chord Transition Tool
One sound becomes many. Let its voices drift, clash, and take on the next recording’s colour.
Files stay on your device. No uploads, accounts, or external processing. Reloading clears the project.
1. Outgoing / A
No recording loaded.
Listen first, then mark the last moment of A to keep.
2. Incoming / B
No recording loaded.
Find the chord or moment where B should enter.
3. Shape the journey
Automatic estimates are approximate. Edit notes with octaves (C4, F#3, Bb4), cents (C4 +25c), or frequencies (261.63 Hz). Nothing is snapped to a scale.
Advanced sound controls
Pitch resolves after 80% of the transition, then holds. Early B overlap can introduce its audio before resolution. Repeated target voices share a note if voice count exceeds chord size.
Timbre Morph — source → synth → destination
Capture sections / stable voiced subsection
Capture positions follow the transition markers until edited. For a noisy scream, choose a voiced subsection here and enter its pitch manually above. Pitch confidence never blocks the effect. Source noise uses an approximate harmonic mask; 100% preserves the recorded noise.
The source is sustained with overlapping grains. A shared, continuous phase clock drives each sampled voice and both harmonic timbres. B matching estimates the mix’s harmonic colour, not separate instruments. Uncertain estimates retain the synth and fade into actual B. Presets also adjust duration and reverb; the sound controls remain editable.
Timbre matching is estimated locally when you preview or render.
Chart strip: source (copper) → synth (blue) → destination (green).
Load both recordings to see the complete timeline.
4. Listen, then export
Import A and B to begin.
WAV includes A up to its marker, the entire Dream Chord, then B from its marker to the end. Video pauses both recordings at their marked frames and dissolves between them during the chord, then resumes B. Video export records in real time: keep this tab visible and the device awake. Preview is the section around the transition; exports contain the complete result.
Limits, formats, and how it works
Use files your browser can decode (typically WAV, MP3, AAC/MP4, Ogg, FLAC or WebM, depending on browser). A video must have a decodable audio track. Unsupported codecs and silent videos produce an error; try converting locally. No microphone or camera access is requested.
Each file is limited to 128 MiB. Decoded audio is limited to 192 MiB combined; audio rendering uses a separate 160–384 MiB working budget (lower on touch / low-memory devices). On memory-constrained devices use short, trimmed recordings and close other tabs. WAV is stereo 44.1 kHz / 16-bit PCM. Video is fitted without stretching into a frame up to 960 pixels wide (30 fps), recorded with the browser's supported WebM or MP4 codec. Audio-only export never requires a video encoder.
Pitch analysis combines periodicity estimation and a harmonic spectrum model in a worker. Chord notes and octaves remain estimates. Actual source grains, a soft harmonic synthesizer and a measured destination harmonic profile share each voice’s continuous pitch path. Source dominates the beginning, synth mediates the middle, and destination colour emerges near the end. No instrument isolation is promised. The intermediate dissonance is deliberate. Smooth envelopes, a gentle limiter and optional reverb soften amplitude edges, without correcting pitches. Preview and offline WAV rendering share the same synthesis graph.
Mobile browsers may require a tap to start sound and may suspend processing when locked or backgrounded. Video export cancels if the tab becomes hidden. Heavy device load may drop frames or slow real-time capture; use WAV for precise audio timing. The short timbre segment is prepared in a worker before playback; repeated previews reuse it. There is no project autosave; save your export before closing.