AI & Creative Tools

Sonilo's Music 1.5 Lands in ComfyUI: Write Lyrics, Get a Sung Track Back

The video-focused AI audio startup's lyrics-to-song model is now a native ComfyUI node — describe a style, drop in a verse and chorus, and the node sings it.

ComfyUI’s November 4 release, v0.44.0, quietly added one of the more unusual partner nodes yet: Sonilo Music 1.5, a lyrics-to-song model from a startup whose whole pitch, until now, has been that you shouldn’t need to write a music prompt at all.

From “no prompts” to a prompt-and-lyrics node

Sonilo made its name on video-native scoring — point its model at a clip and it composes music timed to the footage’s pacing and emotion, no text description required. That’s the product behind its $11 million funding round (B Capital leading, Redpoint participating), announced earlier this year, and the reason its founding team — CEO Shawn Song, CTO Alex Yin, and Keli Li, all TikTok veterans — built it in the first place.

Music 1.5, the node that just shipped in ComfyUI, is a different surface on the same underlying audio model: standalone text-to-song. You write a style description, add a Lyrics: block with your own verse and chorus, and the node sings those words back as a finished track. Skip the lyrics block and it writes its own; skip it along with any vocal language in your prompt and say “instrumental” or “no vocals” instead, and it returns a music bed with no voice at all. Length runs on auto (fits the track to however long the lyrics take to sing) or a custom duration from 10 to 360 seconds, and the combined prompt-plus-lyrics text is capped at 2,000 characters.

Why a ComfyUI node, not just an app

Sonilo already had distribution: it launched a ComfyUI integration back in April, then added a model on fal.ai in June, plus an API layer aimed at video platforms. Native partner-node status in ComfyUI’s own template library (search “Sonilo Music 1.5” after updating) is a step up from a community integration — it means the node ships with Comfy’s own workflow templates rather than requiring a separate custom-node install, putting a funded, video-first AI audio company’s vocal model directly into the same local pipelines artists already use for image and video generation.

It’s also notable for what it isn’t: most of the other audio tools competing for attention in this space are web apps or APIs. A ComfyUI node means Music 1.5 can sit in a graph next to an image or video generator and get triggered as one step in a larger automated pipeline, rather than as a separate destination a workflow has to export out to.