Player not loading? Watch on YouTube
This ComfyUI tutorial uses Pixaroma nodes to turn a still image and an audio file into a music-reactive video. It starts with updates to Easy Installer and ComfyUI, then demonstrates Preview Image. That node lets users inspect results before saving them to the output folder or another location, and pass images into further processing nodes.
Audio React accepts one image and one audio source at a time. Its controls cover motion intensity, overlays, audio bands, output size and frame rate. The speaker explains that the effect uses code for simple 2D transforms rather than an AI animation model. A separate node exports MP4 files. The speaker reports successful 4K output on a 6 GB VRAM card, but tested audio clips under a minute and leaves memory requirements for longer clips uncertain. A CapCut example combines 12 generated clips using the same audio.
The final section sets up ACE-Step for local AI music generation. It requires four downloads: a main model, two text encoders and a VAE. The walkthrough adjusts duration, seed, BPM, key scale and lyrics, then compares a song with Suno AI using the same prompt and words. The speaker reports roughly 20 to 30 seconds for a two-minute song on a 24 GB VRAM card. Results vary by seed and style; the examples include missed lyrics and unclear vocals. The speaker prefers Suno AI's results.
