Have a ComfyUI workflow? Get a readable Workflow Doc in your browser.Make a workflow readable.Try it now ↗
Model Weights : Updated core model and codec files to the latest versions (20260124) for improved adherence and stability.
Runtime profile
Source description
Visit the below link for model downloads.
https://github.com/monnky/ComfyUI-RT-HeartMuLa
Changelog
[1.09] - 2026-01-24
Updated
Model Weights : Updated core model and codec files to the latest versions (20260124) for improved adherence and stability.
[1.06]
Fixed
AI-generated commentary
AI-generated explanation based on source and configuration details. Suggestions are clearly labeled.
ComfyUI RT HeartMuLa Open Source Song Generator is a workflow for creating songs from lowercase, comma-separated tags and producing audio.
The workflow is described for song generation with audio, including long-form tracks of four minutes.
Download version v1.09 from https://civitai.com/models/2329726?modelVersionId=2624766; the listed publisher is Raretutor.
For model downloads, the description points to https://github.com/monnky/ComfyUI-RT-HeartMuLa.
Before using it, verify that ComfyUI-Custom-Scripts is available at pin 1.2.5.
Suggestion · not verified
Before using it, verify that ComfyUI-QualityOfLifeSuit_Omar92 is available at pin f09d10dea0afbd3984a284acf8f0913a634e36ec.
Suggestion · not verified
Before using it, verify that ComfyUI-RT-HeartMuLa is available at pin b188250cfd2291a836da9cc860a5980c5dac3126.
Suggestion · not verified
Format tags as lowercase, comma-separated values.
Suggestion · not verified
The v1.09 description lists split codec support for dual precision between the model and decoder.
The changelog records a fix for switching CFG Scale between 1.0 and higher values.
The description reports improved audio clarity for four-minute tracks; verify that result on your own setup.
Suggestion · not verified
A license is not listed; verify usage terms before use.
Suggestion · not verified
Does this need to be edited?
Sign in to send an edit request.
Sources
1 sourceSource excerpts
2 excerptsSource context: 288 downloads · Type Workflows · Base model Other
CFG Scale Crash : Fixed a critical "Batch Size Mismatch" error when switching between CFG 1.0 and higher values. Implemented a hard KV-Cache reset to synchronize batch dimensions.
Decoder Hang : Fixed an issue where the fp32 codec would hang indefinitely when used with quantized models. Added explicit device casting and eval mode synchronization.
Linting Errors : Resolved formatting issues in the generation logic for better code stability.
Optimized
Long-Form Audio : Improved audio clarity for 4-minute tracks by implementing Temperature Decay and aggressive VRAM cache cleanup to prevent robotic artifacts.
Tag Compliance : Updated conditioning nodes to strictly enforce lowercase and comma-separated tag formatting as required by the paper.
[1.1]
Added
Split codec support (Dual Precision for Model/Decoder)
[1.0]
Initial Release
First public release
Estimated VRAM requirement
0 GB estimated
Requirements
3 requirementsNode pack · Registry
Node pack · Registry
Node pack · Registry
Source context: 273 downloads · Type Workflows · Base model Other
