> ## Documentation Index > Fetch the complete documentation index at: https://docs.youka.io/llms.txt > Use this file to discover all available pages before exploring further. # Choose a Lyrics Sync Model > Compare lyrics alignment and transcription models in Youka Online and Youka Desktop Lyrics sync models do one of two jobs: * **Alignment** keeps lyrics you provide and finds when each word is sung. * **Transcription** listens to the performance, writes the lyrics, and times the words. Choose the job first. A high-quality transcription model can still change lyric text, while an alignment model is designed to preserve it. ## Quick picker | Your priority | Start with | | --------------------------------------------------------- | ---------------------------------- | | Preserve accurate lyrics you already have | **AudioShake alignment** | | Detect lyrics with the best overall accuracy | **AudioShake transcription** | | Try a lower-cost, multilingual transcription option | **ElevenLabs Scribe** | | Try another premium transcription option for clear vocals | **MusicAI transcription** | | Fine-tune timing after any automatic result | [Manual Sync](/en/web/manual-sync) | ## Transcription models ### AudioShake AudioShake is the recommended default for lyrics detection in most songs. The app marks it **Best** and shows a medium cost tier. ### ElevenLabs Scribe ElevenLabs Scribe creates word-timed lyrics from the original performance. The app shows four quality dots and a low cost tier. It supports Auto-detect and a broad set of language hints. If you supply lyrics, Scribe uses them to improve vocabulary recognition. It does not align that text word for word, so review the resulting words and line grouping. Use an alignment model when the supplied text must remain unchanged. ### MusicAI MusicAI is another transcription option for songs with clear vocals. The app shows four quality dots and a high cost tier. The live Lyrics Sync Model menu with ElevenLabs Scribe selected ## Alignment models Use alignment when you already have the correct lyrics and want Youka to place those words on the timeline. * **AudioShake alignment** is the recommended starting point for most songs. * **MusicAI alignment** provides alternative word-level or subword-level timing when available in the selected workflow. ## Language compatibility Keep **Auto-detect** when you are unsure of the language. If an automatically selected model cannot use an explicit language hint, Youka may switch to a compatible model such as ElevenLabs Scribe. If you explicitly selected a model, Youka asks before replacing that choice. ## Where to choose the model Open **Advanced Settings** on **Create Karaoke Video** or **Create Lyric Video**, then choose **Lyrics Sync Model**. The available list changes based on whether you provide lyrics, transcribe from audio, or create a lyric video with the original audio. The same workflow is available in Youka Online and the new Youka Desktop.