How to make a transcription

Always edit a transcription of the lyrics for the AI timing if our portal can find you the song lyrics. The AI can figure out the lyrics but the line spacing and formatting will be wrong, and there will be spelling errors which results in using more time in the editor.

Click "Fetch lyrics" and check the spelling from various sources by clicking next or previous if you're unsure. The portals first suggestion might not be 100% correct.

Fix any grammar issues and change the lyrics if they don't match the audio.

⬇️ If the portal can't find lyrics for you, scroll down this page for “ALIGN WITHOUT TRANSCRIPTION” -instructions.

A good way to make the transcription is to listen through the song once in the audio preview as you edit.

Delete all “chorus”, “verse”, “instrumental break” etc words, we only need the actual lyrics.

Find and remove all punctuation marks etc. such as . , ? ! “ * ‘ ( ) by pressing the "REMOVE UNWANTED CHARACTERS"-button.

Make sure there are no empty lines or other extra markings of any kind.

For line lengths we are aiming for musical phrases that preferably do not have over 25 characters per line (Latin alphabets), but to keep the lyrics singable on screen we must take account that quick lines (that occur for example in rap music) should be longer or they will colour out too quickly on the karaoke screen.

The text view in the editor will now show the 25 character mark limit as a red letter, so you can visually catch lines that are too long. This is simply a visual guide and not an absolute limit - keep dividing the lines musically and as evenly as possible.

Character limits for other alphabets - (same principle applies as above for very fast sections):

Chinese: 11 (no spaces between characters)

Korean: 14 (including spaces)

Japanese: 12 (including spaces)

A good way to determine the line breaks is to listen to the singer, where does the musical phrase or melody end, where do they breathe etc.

If there is a word where one syllable or part of the word is sung noticeably longer than the others, you can use “-” as a way to divide the word into syllables for the AI. For example Disneys Frozen:

Let it go

let it go

can’t hold it back any-more

Please note that words that are pronounced as one syllable such as “have” or “save” do not need to be divided even if they are sung long.

If there is a word with an actual needed “-” (for example: like-minded) please insert a double “- -”, this way the AI will remove the first “-” and leave the second.

Divide into syllables and hyphenate according to the hyphenation rules of your language.

Still feel unsure? Check out another example in the Q & A section!

In case of numbers in the lyrics, do avoid using numerals if/when possible. That being said, one should use their own consideration and language knowledge and think whichever works better for the rhythm and painting of words on screen.

If the song name has a numeral, then use the same style in the lyrics.

Be extra careful with the timing, the AI usually gets numerals wrong and will require fixing.

Capital letters are used only in the beginning of the lines, and if possible we want to retain a sense of complete sentences in the lyrics. For example if you must break one line into three, only the first line will need a capital letter. Example below:

Of course use capital letters, even in the middle of sentences, when it is grammatically correct.

No need to flag the song if the song with the slurs is not racist in itself (it's a song by a POC artist for example).

Song should be flagged if the song with the slurs is by a white artist with racist message or undertones.

Regarding yells, shouts, “oh yeahs” and additional talking (spoken, not sung parts) in the beginning or end of the song, do use your own judgement whether its a vital part of the lead vocal and karaoke singing experience. Also use spelling in the yells that are associated with the current lyrics language, for example is it “Ay” or “Ey” or “Yea” or “Jou”.

If the singer includes their own name or names of the other artists on the track, like a signature, please leave this in. We want to give karaoke singers the chance to be the artist. Our job should not be making creative decisions but a transcription of how that artist performs the song. E.g. DJ Khaled, Jason Derulo songs often feature this.

Do not extend shouts with extra letters. For example don’t write out a long note “Ooooh”, instead write a regular “Ooh” with a longer region length in the waveform view.

In case there are loud backing vocals in the song, or a call and answer kind of thing, usually its easier to write the backing vocals to the lyric transcription, so that the AI will time the actual lead vocals lyrics correctly. You can then remove the backing vocals lyrics in the editor.

If there are no lyrics available in the portal when you press "Fetch lyrics" or the lyrics don't match the song, start the task normally but leave the transcription bit empty.

💪🏻 The AI will guess the lyrics for you.

Check the box "Show lyrics after recognition" and press "Align without transcription". AI guessed lyrics appear after a little while on the next page where you can press "Copy to clipboard".

Then press "Start over" paste it to the transcription page, and edit it like you would normally do a transcription.

In short:

Japanese characters

English + (JP text)

-> make the karaoke lyrics in Japanese characters + insert Furigana where needed.

Furigana can be written above a word by writing it it the transcription after the character or word in brackets. Example: 振(ふ)り仮(が)名(な)

You can also insert Furiganas in the editor by splitting the Kanji from the word into it's own syllable/region using the split tool S, and writing the Furigana in brackets after the Kanji inside the same region.

Example of Furigana in karaoke lyrics:

If the song name is written in

Example of romanized Japanese karaoke lyrics: