Ziyouliangji aims to use AI music platform Hitto to turn everyone into a song creator

Ziyouliangji Information Technology is betting that AI can democratize music creation by moving beyond general-purpose models to tackle the high-barrier music sector. Its platform, Hitto, uses a proprietary, self-developed music foundation model to allow anyone to generate complete songs from text, images, or emotions, aiming to transform music from a professional skill into a mainstream form of emotional expression for ordinary consumers.

Vertical Focus as a Differentiator in a Crowded AI Market

While many AI startups chase parameter counts, Ziyouliangji has deliberately chosen the complex vertical of music. This focus addresses a specific technical gap: general-purpose models often struggle with the unique demands of music, especially for the Chinese market. Music generation involves ultra-long context, melodic structure, and the subtleties of tonal languages like Mandarin, where elements such as soft pronunciations are critical for lyrical coherence. The company posits that this complexity is why many international AI music models haven't successfully adapted locally. By building a music-specific foundation model from the ground up, they aim to create a deeper competitive moat than what a generic large model could offer.

Technical Architecture: Balancing Coherence and Detail

Hitto's technical approach is engineered to solve core problems in AI music generation. The team employs a hybrid Autoregressive + Non-Autoregressive (AR+NAR) architecture. This design allows the model to generate the overall coherent song structure (the AR component) while also producing refined, detailed local elements like specific vocal inflections (the NAR component). The model also operates with multimodal capabilities, understanding inputs across text, image, audio, and video in a unified representation space. This technical foundation is critical for enabling the core user promise: turning a photo or a described emotion into a musically and emotionally coherent song.

From Consumer Novelty to Emotional Utility and Commercial Viability

The platform’s adoption reveals a shift in how music is perceived. User examples—a truck driver turning a poem into a song, families documenting growth through photo-generated music—show music becoming a tool for personal emotional documentation and expression rather than just entertainment. This aligns with the company’s stated goal of "music democratization." Beyond consumer use, Hitto is penetrating commercial and institutional spaces. Its involvement in generating the theme song for the 2025 World AI Conference demonstrates professional-grade capability, while explorations into education, healthcare, and mental wellness suggest a strategic play to position AI music as a functional tool for therapeutic and emotional support, expanding its value proposition far beyond novelty.

The Challenge of Moving Beyond Technical Competence to Emotional Resonance

A key industry insight from the article is the changing battleground for AI music. Competition is shifting from the basic question of "can AI generate a song?" to the more challenging "can that song genuinely resonate with a listener?" Ziyouliangji’s focus on this phase is evident in its efforts to solve a common AI flaw: creating songs that are technically smooth but lack memorable hooks. The V3.0 update specifically targets improved vocal performance, introducing nuanced techniques like humming and breathy vocals, and optimizing melody to be "catchy" and emotionally engaging. This signals that the next frontier for AI music isn't just generation—it's emotional authenticity and artistic appeal, a far more subjective and difficult standard to engineer.