How to separate vocals: two tracks or six stems?
Choose vocals and accompaniment when you need a lead vocal or backing track. Choose six stems when you need separate instrument groups for editing. Start with the best original file you have, then inspect both the extracted track and what was removed; separation can leave audible artifacts.
1. Choose the output for your task
Vocals and accompaniment produces two useful groups for practice, an isolated vocal or a backing track. Karaoke is another backing-track workflow. Six-stem separation gives vocals, drums, bass, guitar, piano and other sounds, which is useful when you want to adjust one group in an editor.
Six stems is not automatically a cleaner vocal result. Choose it when you need the extra instrument groups. Dereverberation targets reverberation, while denoising targets noise; neither is a substitute for choosing the separation output you actually need.
2. Prepare a file and check the limits
Each upload must be no larger than 50 MB and no longer than five minutes. A filename does not prove a file’s duration or format, so check the actual recording. Upload audio you own or are allowed to process. For a longer recording, edit the relevant excerpt locally first.
Use the best source available and avoid repeatedly exporting a compressed file. MP3 uses lossy compression, as explained in the Audacity manual linked below. Converting an existing MP3 to WAV cannot restore information already discarded by compression. Keep your original file while judging the result.
3. Upload, compare and download
- Open the separation page, sign in with Google and choose the appropriate task.
- Upload the file, check its detected duration and the displayed credit estimate, then submit once.
- After processing, listen to a vocal-heavy passage and a quieter passage. Listen to both the isolated vocal and the accompaniment at a comfortable level.
- Check for residual vocals, missing consonants, metallic tails, displaced percussion or pumping. Pick the output that works for your edit rather than assuming every stem is a clean original recording.
- Download MP3 for a compact listening copy or WAV for further editing. Import related stems at the same start position and check alignment in your editor.
4. Calculate the duration-based cost
Vocals and accompaniment or karaoke costs 10 credits per minute; six stems costs 20; dereverberation or denoising costs 8. Charge = max(one-minute rate, ceil(actual duration in seconds × rate ÷ 60)). This rounds credits up, not the whole duration up to a full minute.
For 90 seconds, vocals costs 15 credits, six stems 30, and denoising 12. For a 30-second clip the one-minute minimum applies: 10, 20 or 8 credits respectively. A five-minute file costs 50, 100 or 40. The credit and storage guide has the full comparison table.
5. Keep expectations and a local copy
Overlapping voices and instruments, heavy effects, distortion and compressed sources can make separation harder. Some leakage or changes in tone can remain. This guide does not claim that a mix can always be restored to its original recording stems.
Uploaded audio and processed results are stored for 15 days and then automatically deleted. Download every file you want to keep; buying more credits does not extend a file’s storage period. Payment and credit records have a separate retention purpose and are not removed with the audio.
FAQ
Will WAV make an already compressed recording lossless again?
No. A WAV export can avoid another lossy encoding step, but it cannot recreate audio information lost before upload.
Should I choose six stems just to get a vocal?
Start with vocals and accompaniment if those are the only outputs you need. Choose six stems when the instrument groups are useful for your work.
Sources and related tools
- Vocal Separation
- Credit Pricing & USD Top-Ups
- Audacity manual — Digital Audio Fundamentals (compression and audio quality)
Check the current product page and task quote before submitting. Tutorial examples illustrate the workflow and do not guarantee results.