🎼 Text2midi-InferAlign
Improving Symbolic Music Generation with Inference-Time Alignment
Welcome to the demo page for Text2midi-InferAlign!
Our method improves symbolic music generation by optimizing inference-time alignment between captions and midi outputs.
Check out the examples below 👇
🎵 Examples
Each example shows:
- Input Caption
- Generated midi (rendered to mp3, playable in browser)
Example 1
Caption: “A cheerful and melodic pop Christmas song featuring piano, acoustic guitar, vibraphone, bass, and drums, set in the key of Eb minor with a fast tempo of 123 bpm and a 4/4 time signature, creating a joyful and relaxing atmosphere.”
Example 2
Caption: “A melodic electronic song with ambient elements, featuring piano, acoustic guitar, alto saxophone, string ensemble, and electric bass. Set in G minor with a 4/4 time signature, it moves at a lively Presto tempo. The composition evokes a blend of relaxation and darkness, with hints of happiness and a meditative quality.”
Example 3
Caption: “This motivational electronic and pop song features a clean electric guitar, rock organ, synth voice, acoustic guitar, and vibraphone, creating a melodic and uplifting atmosphere. Set in the key of G# minor with a 4/4 time signature, the track moves at an energetic Allegro tempo of 120 beats per minute. The chord progression of Bbm7 and F# adds to the song’s inspiring and corporate feel.”
Example 4
Caption: “This short electronic song in C minor features a brass section, string ensemble, tenor saxophone, clean electric guitar, and slap bass, creating a melodic and slightly dark atmosphere. With a tempo of 124 BPM (Allegro) and a 4/4 time signature, the track incorporates a chord progression of C7/E, Eb6, and Bbm6, adding a touch of corporate and motivational vibes to the overall composition.”
Example 5
Caption: “An energetic and melodic electronic trance track with a space and retro vibe, featuring drums, distortion guitar, flute, synth bass, and slap bass. Set in A minor with a fast tempo of 138 BPM, the song maintains a 4/4 time signature throughout its duration.”