3 Reference Slots for Voice or Music
Add a voice sample, a music bed, and a sound effect in any mix of up to 3 clips. Tell the prompt which clip shapes which part. That keeps each source in its own role.
Create Audio from Audio Now

Transform existing audio into new sounds with MagicLight AI. Upload a voice recording, music, or sound effect, describe the changes you want, and generate a new audio clip. No editing skills needed.
Create Audio from Audio Now
Pick the cleanest clips you have before you upload. Magiclight AI Audio to Audio Converter then gives each reference a clear place in the new sound.
Add a voice sample, a music bed, and a sound effect in any mix of up to 3 clips. Tell the prompt which clip shapes which part. That keeps each source in its own role.
Create Audio from Audio Now
MP3, WAV, PCM, OGG, and Opus are all supported in Magiclight AI. A phone voice note in Opus can sit beside a WAV from a field recorder. No studio mic is needed for a usable clip.
Create Audio from Audio Now
Suno cannot read reference audio, so this mode runs on Seed 1.0 alone. The model reads every reference you add, without needing you to compare a long list before you start.
Create Audio from Audio Now
Say what the new clip should hold in everyday words. Try “use the narrator voice to say doors open at nine over the same soft piano bed.” One clear idea works best in Magiclight AI.
Create Audio from Audio Now




Cut each reference to its cleanest few seconds first. In Magiclight AI, follow 3 simple steps, then carry Audio to Audio from the mode switch to the finished clip.
Click Audio in the Toolbox and set mode to Audio to Audio. Seed 1.0 is the only model offered here. Clip the cleanest few seconds of each reference before you continue.
Add up to 3 files in MP3, WAV, PCM, OGG, or Opus (20MB or less). Play each one in the panel to confirm the take. Put the most important clip first and say what each clip contributes in 500 characters or less.
Choose the Audio format and click Generate. Play the result and download it when it sounds right. If one detail is off, adjust the note and run it again.
Podcaster
I uploaded last season’s intro and the host voice sample and wrote one line. The new intro came back close to the old one. I downloaded the WAV and used it that day.
Music Producer
I use a voice clip and a pad from my own library as references. Naming each clip’s job in the note kept the result tight. Re-edit saved me from re-uploading everything.
Sound Designer
Seed 1.0 reads every reference, so I add a door slam and write one change, such as heavier. I run it a few times and pick the version that fits the pack.
YouTuber
I fixed a flubbed line by uploading a clean stretch of my voice and typing the corrected sentence. I checked the clip in the panel first, and the result matched my take.
No, Magiclight AI Audio to Audio builds new audio from your references and leaves your original file unchanged. The tool does not remove noise or cut sections.
Only with that person’s clear permission and ideally in writing. Magiclight AI is meant for voices you have the right to use, so keep proof of approval.
Magiclight AI may stop that run with a copyright error and return no audio. Stick to your own tracks, licensed music, and sounds you recorded yourself. Those references give the cleanest and safest results.
Suno does not take reference audio, so Magiclight AI shows only Seed 1.0 in this mode. For music from a written prompt alone, go back to the Text mode where Suno is the default choice.
Yes, Audio to Audio is a member feature like the other Audio tools, so free accounts cannot use it. Check the plan page for what each tier includes. Commercial use needs a plan with commercial rights.



The voice or beat you need may already sit in your recordings. Upload a clip to Audio to Audio and describe the new version.