New BEST local AI music generator is here!
For AI filmmakers needing custom score generation, YuE2 offers free, local audio creation and flexible reference-based music re-harmonization without relying on cloud subscription models.
What this lesson covers
This video introduces YuE2, an open-source local AI music generation model that runs in ComfyUI with low VRAM requirements. The creator showcases its ability to generate multi-lingual music, perform genre covers from reference audio, adapt scores via AI agents, and provides setup instructions.
Key takeaways from the creator
AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.
- 00:00 ↗
YuE2 is an open-source local AI music generator that can execute on 4 GB of VRAM or less to create songs and covers.
- 05:19 ↗
YuE2 can process reference audio files to preserve underlying melodies while altering musical style, key, or lyrics.
- 07:55 ↗
Connecting YuE2 to language agents enables iterative audio editing and prompt-based song adjustments.
- 09:21 ↗
The model operates by drafting an editable score for vocal and instrumental tracks prior to rendering the full audio file.
- 12:28 ↗
Local deployment uses ComfyUI by updating the installation and dragging in the YuE2 workflow JSON file.
Workflow outlined in the video
- Update ComfyUI to the latest build using the update script in the ComfyUI root directory.
- Download the YuE2 model weights from Hugging Face and the official workflow JSON file.
- Load the YuE2 workflow JSON directly into the ComfyUI interface.
- Provide text descriptions for music style and lyrics, or upload a reference audio clip for cover generation.
- Run the generation pipeline in ComfyUI to synthesize the generated audio output.
Before you use this workflow
These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.
ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.
Explore the tools
How to interpret the numbers
111,837 views captured 16 September 2026. ReelStack recommendation score: 59/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.
Show the calculation
- 65% performance: 0.66× current views vs the channel's recent median; same-age history is not yet available. View-evidence factor 99% (views / (views + 1,000))
- 25% freshness: 94/100 with a 14-day half-life
- Momentum pending: collecting daily snapshots; its 20% weight goes to observed performance, not free points
- 10% engagement: 100/100; likes + 4× comments, smoothed with a 500-view neutral prior