How to prepare your own audio for an AI rap video
Prepare a short MP3 or WAV, match it to the video duration, and understand audio-reference behavior before uploading. A practical checklist based on our actual upload rules and sample workflow.
Availability note: these are completed controlled tests, not a list of live services. Public generation and purchases are currently closed. The studio and pricing page show current availability.
1. Pick a short section you are allowed to use
Use your own recording or audio you have explicit permission to use. Prepare the intended section in your audio editor before uploading: this studio is not a full-song timeline editor. For our controlled tests, we reused a 10.08-second original track from our first generated sample. This describes the test material, not a guarantee about ownership of every possible generated output.
2. Export MP3 or WAV within the actual limits
The upload validator accepts MP3 (audio/mpeg) and WAV (audio/wav), up to 15 MiB (15,728,640 bytes). Prepare a clip between 4 and 30 seconds; the validator has a small 30.2-second measurement tolerance, not a promise of longer supported videos. Renaming an MP4, M4A or another file to .wav does not convert it. Use a genuine audio export and check that it plays locally.
3. Match the clip to the requested video
The audio must cover the selected video duration. A 10-second request needs at least 10 seconds of valid audio; a short clip cannot fill a longer request automatically. The six examples in our comparison use a requested 10 seconds at 720p and 9:16. The 30-second upload ceiling does not prove that every model supports 30-second output. Choose only a duration offered by an available model.
4. Make the performance easy to follow
Listen for clipped peaks, long silence, abrupt cuts and a beat that overwhelms the voice. Leave enough of the phrase to make the ending understandable. Clean recording is useful preparation, but it cannot guarantee lip sync or musical quality. The studio does not promise automatic mastering, exact beat synchronization, voice cloning or a separate song export.
5. Select the audio workflow only when available
When generation opens, sign in, choose Use my audio, select an available compatible model, and upload the prepared clip through the studio. Match the portrait layout and video settings, then review the server credit quote before submitting. If uploads or a model are unavailable, do not treat a visible control or a completed sample as permission to generate. Public creation is currently closed.
6. Resolve common failures before trying again
Unsupported format: export a real MP3 or WAV. Too large: shorten or re-export below 15 MiB without merely changing the filename. Too short: prepare a clip that covers the selected duration. Unavailable model or missing quote: wait for a supported combination instead of changing file metadata. An audio reference may be reinterpreted in the output; review the finished sound and mouth movement rather than assuming the original track was preserved.