Add Instrumental to Vocals and develop your song idea

To add instrumental to vocals, start with an a cappella recording or hummed melody and describe the music you want. Prepare and listen to the source before generating. Results may change the voice or performance; generation requires login and at least 20 usable credits.

1Source audio

MP3, WAV, M4A, AAC, OGG or FLAC · up to 50 MB · 5 seconds–8 minutes

2Music style

Song title

Advanced options

Lyrics (optional)Avoid these elements

The generated song may change your voice or performance. It does not guarantee an unchanged recording.

Loading…

Upload a vocal and describe its style. Your songs will appear here.

Add Instrumental to Vocals: build a backing idea around your recording

Use the recording to define the musical starting point

A voice memo can carry a melody before you have chosen its instruments. Start with the verse, refrain or hummed phrase you want to develop, then decide what the music should do around it. A soft delivery may call for a quiet acoustic backing; a rhythmic phrase may benefit from a clear pulse. Upload one recording and describe that direction to create a full song draft. Choosing a specific musical starting point gives you something concrete to listen for when the versions are ready.

Try a simple arrangement brief for your first version: acoustic guitar supporting the verses, light percussion, and more space before the refrain. For a stronger groove, ask for a steady beat and a bass line that supports the vocal rhythm. Keep the source recording as your reference and compare how each version develops its melody. A finished draft can help you decide which arrangement ideas to carry into your next rehearsal or production session. If you need a vocal from an existing song first, prepare it with the vocal-separation tool.

Prepare one clear take that is easy to compare

Choose one take whose words or hummed notes are easy to follow. Record in a quiet room, keep a steady distance from the microphone, and leave room at the beginning and end of the phrase. Listen for distortion on loud notes and for background sound that masks the melody. Natural breaths and short pauses help you judge the phrasing later. Keep the whole final note when preparing an excerpt. Save the original take alongside your prepared file so you can compare the source consistently across different arrangement requests.

Accepted uploads are MP3, WAV, M4A, AAC, OGG, and FLAC, up to 50 MB and between 5 seconds and 8 minutes. You can also choose audio available in your account's cloud library. A correct extension alone is not enough; the server checks the file content and decodes the audio. Use the source player to confirm it is readable before submitting. There is no clip-selection editor inside this workbench, so prepare a shorter excerpt elsewhere if necessary. A complete phrase with recognizable beginnings and endings is easier to review than unrelated takes stitched together.

Describe the role of the instruments, not just a genre

A style preset is an editable starting point. Acoustic pop, jazz piano, a rock band, electronic music, lo-fi and hip-hop each fill the style field with a direction you can refine. Choose the closest one, then explain the instruments' roles: piano carrying the harmony, guitar answering between sung lines, or drums giving a gentle pulse. Say where the arrangement should be sparse and where you want more energy. Read the populated field again after changing a preset, because the new selection replaces the previous text.

Write a brief you can evaluate by listening. For an intimate song, try warm piano, acoustic guitar between phrases, soft drums and a quiet opening. For a brighter version, ask for a steady pop groove, a clear bass line and a fuller refrain. Keep one main style and add a few details about texture or energy. Words such as relaxed, brisk, sparse or spacious express useful musical intentions. Exact tempo and key are requests in the text rather than dedicated editing controls, so compare the generated draft with the feel you intended.

Keep the style, title, lyrics, and exclusions separate

The style field is required and accepts up to 1,000 characters. The song title is optional, up to 80 characters; a blank title uses the source file name as a starting label. Put arrangement directions in the style field rather than treating the title as another prompt. In advanced options, lyrics are optional and can contain up to 5,000 characters. Use that field for words relevant to your recording, not a list of instruments. Leaving it empty is allowed. These controls shape a generation request; they are not a text-alignment editor for correcting individual words in the existing audio.

Use the exclusions field for one or two sounds that would distract from this version, such as heavy drums or distorted guitar. Keep those choices consistent with the main arrangement. For a gentle acoustic version, a short exclusion list helps make the direction clear; for a rock version, specify the guitar character you want to retain. Read the style and exclusions together before submitting. If you are also changing lyrics, note which phrases are new so you can compare the performance and arrangement as separate creative decisions.

Compare the draft against the same listening goals

A successful request may provide one or two audio results; the number is not guaranteed. Listen to the source and each available version before choosing a favorite. Pay attention to whether the accompaniment leaves room for the phrase, whether a drum pattern helps the intended feeling, and whether later sections still fit the initial idea. Check the whole result instead of comparing only the loudest moment. A completed task means audio is available, not that every arrangement instruction was followed. The output length also needs checking rather than assuming it precisely reproduces the source duration.

For another version, change one decision at a time: lighten the rhythm, reduce melodic layers, or replace the supporting instrument. Keep the same source take and write down the style used for each attempt. This makes it easier to hear whether the new choice helps the words or the melody. Download useful song drafts as MP3 and give them names that identify the recording and arrangement. You can reopen available results from account history, while keeping important files in your own project folder for later rehearsal, editing or sharing.

Understand the difference between generation and preserving a vocal

A request to create an instrumental for your vocals can suggest keeping the recording's character, but the generated song may alter the voice, phrasing, melody, or performance. The server stores the model's returned audio; it does not replace that audio with a separately mixed copy of your original vocal. Listen especially to pronunciations, held notes, and recognizable vocal details. Neither a familiar voice in one example nor a completed status proves that another recording will remain unchanged. The source player helps you compare what you submitted with the new interpretation.

If the original performance must stay intact, use its recording with an authorized instrumental in a separate mixing workflow. This tool is useful for exploring arrangement ideas and comparing a new interpretation with that source. Its download is a full song draft in MP3; separate accompaniment or MIDI belongs to another workflow. Before publishing, check the rights to the source recording and the applicable terms for the generated song. Decide which vocal details matter to your project and keep the version whose actual sound suits that use.

Check your usable credits before generating

You can choose local audio, play the take and prepare its musical direction before signing in. Each generation costs 20 credits. Once you sign in, any usable account credits can cover the request, including registration gifts and daily check-in credits. Check your balance before submitting; an active subscription or a purchased credit pack is not required. A rejected request or confirmed generation failure returns the deducted credits. Keep your source and style notes when comparing another arrangement, and check the remaining balance before starting the next version.

After completion, available results can be reopened from history, previewed, and downloaded. Save important files yourself and retain notes about the recording and directions. If a submission remains under verification, follow the displayed guidance instead of repeatedly sending it; uncertain acceptance is not the same as a confirmed failure. Do not plan a release around a promised processing time, exact output count, or model quality that has not been checked on your recording. The final review is yours: listen to the full audio, verify suitability and rights, then decide whether the draft is ready for the next stage.

To prepare a complete excerpt before uploading, trim the recording with the audio cutter.

If you want to start from a text idea instead of a vocal file, explore the AI music generator.

To prepare a vocal from an existing song, visit the vocal-separation tool.

Buy 400 credits for $9.90 in a one-time purchase

Questions about giving a vocal musical backing

How do I add accompaniment to a vocal recording?

Upload or select your source audio, play it, and choose or edit a musical style. Add an optional title, lyrics, or exclusions, then review the estimate and account conditions before generating.

Which formats and recording lengths are supported?

MP3, WAV, M4A, AAC, OGG, and FLAC, up to 50 MB and 5 seconds–8 minutes. The file must decode correctly. Prepare an excerpt elsewhere if you need to shorten it.

How should I describe the backing music?

Start with one style and explain the supporting instruments, rhythmic feel, and space around the voice. Presets fill an editable field; they are not separate mixer channels or exact tempo controls.

Do I have to enter a title or lyrics?

No. Style is required, while title, lyrics, and exclusions are optional. A blank title uses the source file name; lyrics can be left empty.

Will my original vocal stay completely unchanged?

Not necessarily. Generation may change the voice or performance, and the server does not mix the original vocal back into the returned song. There is no guaranteed accompaniment-only stem export.

What can I download, and how many credits do I need?

Available song versions can be downloaded as MP3. Sign in and have at least 20 usable credits to generate. Gift, check-in, subscription and purchased credits can all be used; no paid plan is required. Each request costs 20 credits, and confirmed failures return them.