What it does
Turn your MusicXML score into a singing demo in minutes.
Score - Singing Demo
Upload MusicXML and get a realistic vocal preview without a DAW.
Chat-driven Takes
Ask for parts, verses, and style changes using natural language.
Fast Iteration
Generate multiple interpretations quickly for practice or review.
Why SightSinger.app?
Speak music, not MIDI.
| Feature | SightSinger.app | Professional DAW (Digital Audio Workstation) |
|---|---|---|
| Interface | Natural Language (Chat) | Piano roll, parameters, phonemes |
| Learning Curve | Zero | Steep |
| Focus | Global style control, demo quality | Detailed note-level control, production quality |
| Time to result | Minutes | Hours |
How it works
Upload your score
Drop in MusicXML from MuseScore, Logic Pro, Finale, or Sibelius.
We parse tempo maps, part labels, and lyric syllables with music21.
Tell the singer
Use natural language to pick parts/verses and shape phrasing, tone, and expression.
Gemini Flash 3 calls internal MCP tools to map intent into performance parameters.
Generate the singing voice
SightSinger.app runs a custom singing synthesis pipeline to render a realistic vocal demo directly from the score.
Voicebanks are DiffSinger-compatible OpenUtau ONNX models, so adding new voices is straightforward.
Refine and share
Iterate quickly, render new takes, and export a shareable demo for singers.