Suno’s AI music creation tool continues to develop amidst ongoing legal challenges from major record labels. The recently unveiled Suno v5 shows notable technical enhancements compared to its predecessor, v4.5+, particularly in audio quality, with cleaner mixes and improved instrument separation. During a demonstration, product manager Henry Phipps noted that the model identifies isolated sounds and replicates them effectively across the stereo field.
While Suno v5 asserts a better grasp of musical genres, testing indicates mixed results. For example, prompts intended to generate specific styles, such as avant R&B or early ’90s lo-fi indie rock, often yielded downtempo tracks that lacked desired characteristics. Experiences with genre-specific prompts sometimes produced results that did not align with expected sounds from specific eras.
Despite these challenges, the complexity of arrangements in v5 has increased, with more varied song structures that move beyond simple verse-chorus patterns. The model can also remix existing tracks, although it sometimes fails to capture the raw essence of older recordings, producing results that feel inauthentic.
Phipps acknowledged that while the model’s vocal performances have become more nuanced, they still lack the emotional depth that characterizes human performances. Changes in the model’s marketing language have been noted, shifting from terms like “emotionally rich” to descriptions of vocals as “natural” and “authentic.” Feedback on the model’s outputs suggests that, although the vocals are technically improved, they often remain generic and do not fully convey the emotional weight typically present in human artistry.
Source: https://www.theverge.com/ai-artificial-intelligence/786349/suno-ai-music-generator-v5-review