Video timeline:
0-5:55 How to use the editor
5:55-15:52 What to fix in the editor
TIMING ISSUES
Almost always the whole text timing from the widget is about 150-250ms behind, so you can use Cmd + A to select all regions and nudge them a tiny bit forward in the waveform view.
Check the timing of the words, they might be sometimes timed wrong by the AI. Listen and edit the word timing and length by dragging from beginning or end of the regions.
You can check if there are transients/visible consonants in the waveform, they will also give a good indication where the words start.
Be super duper careful with words that start or end with vowels (especially long ones), AI will usually start these notes too late.
Press shift and drag over regions to select and move multiple regions at once. You can also copy-paste one or multiple regions.
WORDS THAT SHOULD BE DIVIDED INTO SYLLABLES:
Sometimes we might forget to hyphenate a word or two in the transcription phase, and notice that words should be divided into separately timed syllables in the editor.
This can be done by setting the cursor where we want the word to be divided and pressing S.
If the song is playing and you press S without stopping the playback, the word will be divided by the cursor and you have to manually fix the divination of the text in the word regions.
If you have stopped the playback and placed the cursor where you want to divide the word and press S, the word editing opens and you can place the insertion point to the point of the word where you want to divide it and press enter. Now it will automatically divide the word into ready syllables with the divided text as well.
Example:
Word we want to divide where the cursor is:
2. After pressing S and the text divination edit opens:
3. Finding the place where I want to divide the word with the insertion point:
4. After pressing enter the region is divided where the red cursor is and word was divided where we placed the insertion point.
MINI REGIONS / GLUE TOOL:
Really short word regions will not color out properly in Singa, and require you to manually make them longer. You can also press shift and drag over regions (one line) to select them and press A to glue the regions together. Alternatively you can select all (ctrl/cmd A) and press G to glue everything, except line breaks. This is a quick way to glue all lines in a song, see below for more.
Example (zoom 150):
Before:
After:
FAST LINES:
When facing really fast lines (most common in Rap and modern Pop), do use the glue tool A to align the short regions together for a smoother outcome + do some manual editing if necessary!
GLUE TOOL:
Press ctrl/cmd A to select all subtitle regions, or select multiple lines yourself:
G — align right: stretches each word's end forward to meet the next word's start.
Shift + G — align left: stretches each word's start backward to meet the previous word's end.
G/shift G will glue everything, except for line breaks (which often mark a gap in the singing). This is the fastest way to do it, but be careful not to glue over actual gaps in between words on one line. You can glue lines together which don't have a gap between them by using the below tool after.
Press shift and drag over to select multiple regions:
A — align right: stretches each word's end forward to meet the next word's start.
Shift + A — align left: stretches each word's start backward to meet the previous word's end.
A/shift A will glue over line breaks, so you can use it to glue multiple lines together in need after using G/shift G. You can also use A/shift A if you just want to glue one specific section and need more fine control over the gaps in between words.
Example (zoom 100-150):
Before:
After:
LINE BREAKS:
If you notice any red letters on the lines, that is the character counter showing you which lines go over the suggested 25 character per line.
Even though you edited the line lengths while making the transcription, do double check the lines with red letters to make the final decision whether the line should be kept as is or is a line break needed.
NOTE: I do want to remind you that the limit is not absolute and faster lines/rap lines still often need to be longer than the 25 character limit suggests.
Keep editing as you have until now -> divide the lines musically and as evenly as possible, this feature is simply here as a visual mark.
Before:
After:
LANGUAGE-SPECIFIC TIPS:
Spanish
Be careful with words ending in vowels followed by words starting with vowels:
va a adivinar
que él
The AI often struggles with a combination of short words and vowel-starts. You can use the glue tool to drag out the short segments and manual adjustment to ensure the words / syllables are starting in the right place where the vowel sound changes to the next word.
Make sure to split "ia", "ie", "io" or "ion" as two syllables, if they are stretched out in the singing (either at the end or in the middle of a word):
dí-a
frus-tra-ci-ón
vol-ve-rí-as
te di-osi-en-to
Think about how the word is sung. If the stretched syllable is the "tra" in the above frus-tra-ci-ón and the ending is quickly as one sound, leave as is. But often, particularly for these kind of endings, they are sung as distinct phonetic sounds, even in some cases where it would be breaking a dipthong grammatically. The subtitles should match how the singer separates the sounds in the word, it is just about the letters colouring out with the right timing.
Split all the syllables in the word for more accuracy, not just the last one. This is easiest to do at transcription stage.
Words might happen to colour out correctly if sung with an even rhythm and when the word is quite symmetrical in width: e.g. por-que but with a word like a-mor, "a" will colour out much more quickly than "mor" since it is just one letter, even if the singer gives equal weight to each syllable. It's not always necessary to split this word (if not noticeable in the video preview), but it's just a good example of why when a word is stretched, some might be fine without splitting the syllables and others might need extra editing. So keep an eye out for this!
The glue tool can be especially useful in Spanish songs, which are often quite fast. Be careful to use Shift A or A appropriately depending on the song - do the segments need to snap back (starting slightly late - often the case for words starting with vowels - Shift A) or be extended out at the ends (they end too early - A)? Make sure to double check that using the glue tool hasn't interfered with the timing particularly regarding the first point above.