MeloDIY

AI Model Disclosure

Last updated: September 19, 2026

This page identifies the models used by the customer-facing AI and machine-learning features in the Service. It also explains their purpose, the inputs they process, and material limitations.

Model names belong to their respective owners. Listing a model does not imply sponsorship, endorsement, or an official partnership.

1. AI music generation

  • Model: music generation model available through kie.ai.
  • Use in the Service: generate instrumental or vocal music from a prompt or user-supplied lyrics.
  • Inputs processed: prompt, selected style, title, instrumental setting, optional lyrics, and generation settings.

2. AI lyrics generation

  • Model: lyrics generation model available through kie.ai.
  • Use in the Service: create a lyrics draft from a topic, mood, structure, style, and language request.
  • Inputs processed: the lyrics prompt and selected creative settings.

3. Vocal removal and stem separation

  • Model: Demucs.
  • Use in the Service: estimate vocals, drums, bass, guitar, piano, and other stems from uploaded or saved audio.
  • Inputs processed: uploaded or saved audio and the selected separation settings.

4. Audio to MIDI

  • Model: Spotify Basic Pitch.
  • Use in the Service: estimate notes and pitch bends and create a MIDI file from audio.
  • Processing location: in your browser. Audio used by this feature is not uploaded to our server unless you separately use another upload feature.
  • This transcription model is optimized for clear pitched material and may be less accurate on dense mixes, percussion-heavy audio, noise, or overlapping instruments.

5. Safety, accuracy, and rights

Models can produce unexpected, inaccurate, offensive, or similar outputs, and source-separation or transcription results are estimates. Model safeguards and our Acceptable Use Policy reduce risk but cannot guarantee that every harmful request or output will be detected.

You must review results before use, avoid relying on them as factual or professional advice, and confirm that your prompts, uploads, lyrics, voices, brands, and final use respect law and third-party rights. Report unsafe or infringing material through the channel in our Acceptable Use Policy.

6. Model and version changes

We may update a model or version to improve quality, safety, or availability. A different model may be used temporarily when a feature is unavailable, and this disclosure will be updated after a material change.

7. Model questions

Contact support if you need clarification about a model used for a specific feature or want to report a model-safety concern.

Support email: support@melodiy.org