Skip to main content
If you don’t have a song’s lyrics on hand, Youka can detect them from the audio. The AI listens to the recording, transcribes the words, and syncs each word to the exact moment it’s sung.

​
When to use auto-transcription

Auto-transcription works best when:
  • You don’t have access to the song’s lyrics
  • The lead vocal is clear and easy to follow
  • The song uses standard pronunciation in its language
  • You want to get a karaoke up and running quickly without sourcing lyrics manually
If you do have the lyrics, providing them manually typically gives better timing accuracy. See Add Lyrics.

​
Steps to use auto-transcription

1

Open the Create Karaoke page

Click Create Karaoke from the Youka home page, then upload your file or paste a URL.
2

Select Detect from audio

In the lyrics section, choose Detect from audio instead of I have lyrics.
3

Choose the AI model (optional)

Open Advanced Settings to select the transcription model. Each model offers a different balance of accuracy and credit cost:
ModelStrengthsCredit cost
AudioShakeBest accuracy for most songsStandard
MusicAIPremium detection with syllable-level timingHigher
WhisperBudget-friendly; good for common languagesLower
If you’re unsure which to choose, AudioShake is a reliable default for most use cases.
4

Select the song's language

Choose the language the lyrics are sung in. This helps the model apply the correct pronunciation patterns.
5

Click Create Karaoke

Click Create Karaoke. The AI will listen to the audio, identify the words, and sync each one to the music. Processing typically takes 2–3 minutes.

​
Credit cost

Auto-transcription uses more credits than providing your own lyrics because the AI performs an extra analysis pass. The exact cost depends on:
  • The duration of the song
  • The AI model you select
The precise credit amount is shown on the creation screen before you confirm.

​
Reviewing transcription results

AI transcription is accurate for most songs, but it can make mistakes. After your karaoke is created, open the project and check for:
  • Misheard words — the AI may transcribe a word phonetically rather than spelling it correctly
  • Names and proper nouns — song-specific references, artist names, and place names are common error points
  • Stylized pronunciations — words that are deliberately mispronounced or altered as part of the song’s style
Use the Studio editor to correct any errors before fine-tuning the timing.
For songs with heavy reverb, background noise, or multi-part harmonies, consider sourcing the lyrics and using the I have lyrics option instead. It gives the AI a precise reference to work from and usually produces tighter timing.

​
What’s next

Edit lyrics

Fix any transcription errors directly in the Studio editor.

Manual sync

Fine-tune the timing of individual words and lines.