Singing synthesis is an important topic for speech communication research, because it touches on several areas, like voice quality, vocal emotion and expression, text-to-speech synthesis, voice personality, speech signal modeling and synthesis, voice transformation, and of course music.
Singing synthesis is particularly attractive and demanding. Sound quality and expressivity are of the utmost importance, with requirements even greater than for speech synthesis. Musical aspects of singing synthesis open new research areas, like interactive or performative synthesis, studio quality sound processing for audiovisual applications.
In addition, a singing synthesis is attractive and fun because it is a bit aside of the mainstream of speech communication research.
The special session “Singing Synthesis Challenge: Fill-In the Gap” aims at bringing together research teams working on singing synthesis from all over the world by means of proposing a common challenge.
The last singing synthesis challenge took place at Interspeech 2007, and we think it timely to renew this very rewarding experience.
The challenge will be to fill-in the gap in a well-known song (e.g., “Autumn leaves”), i.e., to synthesize a new, specially written couplet. The new couplet includes new lyrics, and possibly a new melody, to be inserted in the song. The chosen song is a top hit song, so that a large number of interpretations are available on the web and can be use for reference, machine learning, comparison, etc. In addition, the participant are free to propose any other significant examples of singing synthesis.
The contribution format is as open as possible, all aspects of the challenge and all styles of synthesis methodologies being welcome, e.g., both off-line (studio) singing synthesis systems, with no limits on time for producing the result, and performative (real-time) singing instruments. Papers of singing synthesis evaluation methodology are also encouraged.
Prior to the conference the different singing synthesis contributions will be assembled on the Chanter project web page using the a cappella version of the submission. The different synthesis will be provided anonymously. The page will be publicly announced on different mailing lists and people will be kindly invited to assess the quality of the different songs. The contributing authors will have the opportunity to discuss the detailed assessment methodology. Among the intended criteria of the evaluation are the singing voice quality and the musical quality of the interpretation.
Additionally a public voting procedure will be organized during the Special Session and allowing Interspeech attendants to vote for the “Best scientific contribution to singing synthesis research”.
The winners of the evaluation campaign and the voting procedure will receive an official InterSpeech Singing Synthesis Challenge Award. Additionally, the winning contributions in the different categories will be played at the InterSpeech social event.