pymaster17 commited on
Commit ·
c2ba9d0
1
Parent(s): af100d8
Add musician-friendly demo guide
Browse files
README.md
CHANGED
|
@@ -34,6 +34,49 @@ reference.
|
|
| 34 |
|
| 35 |

|
| 36 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
## How it works
|
| 38 |
|
| 39 |
VocalRender combines three components:
|
|
|
|
| 34 |
|
| 35 |

|
| 36 |
|
| 37 |
+
## For musicians: try it without writing code
|
| 38 |
+
|
| 39 |
+
You do not need to understand machine learning or install the model to try
|
| 40 |
+
VocalRender. Open the **[online VocalRender demo](https://huggingface.co/spaces/pymaster/VocalRender-demo)**
|
| 41 |
+
and work with familiar score formats in your browser.
|
| 42 |
+
|
| 43 |
+
### What VocalRender does
|
| 44 |
+
|
| 45 |
+
VocalRender sings a melody that you provide. You choose the lyrics, notes,
|
| 46 |
+
rhythm, tempo, and a short voice reference; the model renders them as a singing
|
| 47 |
+
voice. It is a score renderer rather than an automatic composer: it does not
|
| 48 |
+
write the melody or arrange an accompaniment for you.
|
| 49 |
+
|
| 50 |
+
### What to prepare
|
| 51 |
+
|
| 52 |
+
- **A melody:** paste ABC notation, upload an ABC file, or upload MusicXML from
|
| 53 |
+
notation software such as MuseScore. You can also enter notes manually.
|
| 54 |
+
- **Chinese lyrics:** embedded ABC `w:` or MusicXML lyrics can be imported. If
|
| 55 |
+
the score has no note-aligned lyrics, type or paste them in the lyrics box.
|
| 56 |
+
- **A voice reference:** select one of the included voices, or upload 2–8
|
| 57 |
+
seconds of clean, unaccompanied singing that you have permission to use. The
|
| 58 |
+
clip supplies the vocal color; it does not need to be the same song.
|
| 59 |
+
|
| 60 |
+
### Score-to-song workflow
|
| 61 |
+
|
| 62 |
+
1. Open **Import ABC notation or MusicXML**, provide your score, and press
|
| 63 |
+
**Parse score**.
|
| 64 |
+
2. Check the suggested vocal part, lyric line, and measure range. The demo
|
| 65 |
+
preselects a short range that fits the model; you can change it.
|
| 66 |
+
3. Press **Load selected range into editor**. Review the lyrics, pitches,
|
| 67 |
+
durations, and BPM, then press **Generate Singing**.
|
| 68 |
+
|
| 69 |
+
In the editor, MIDI pitch `60` means middle C (`C4`), and pitch `0` means a
|
| 70 |
+
rest. **+ Melisma note** lets one lyric syllable continue across multiple
|
| 71 |
+
notes. `SP` represents a rest or breath.
|
| 72 |
+
|
| 73 |
+
The released checkpoint currently works best with Mandarin Chinese and a
|
| 74 |
+
single vocal melody. Chord names in ABC, such as `"C"` or `"G7"`, are treated
|
| 75 |
+
as accompaniment labels and ignored, but actual simultaneous notes are not
|
| 76 |
+
supported. For piano or ensemble scores, select a monophonic vocal/melody part.
|
| 77 |
+
The demo reports unsupported passages instead of silently guessing which note
|
| 78 |
+
should be sung.
|
| 79 |
+
|
| 80 |
## How it works
|
| 81 |
|
| 82 |
VocalRender combines three components:
|