HTDemucs Ft

HTDemucs-ft provides pretrained weights for the htdemucs_ft model, which is designed for music source separation in the waveform domain. The resource is distributed as a collection of four separate models, with their outputs averaged to produce results. Each model's weights are stored in a .safetensors file, which includes the model's class and initialization arguments in its metadata. Additional training metadata is available in corresponding .json sidecar files. The files and their structure are intended for use with the Demucs framework, and further instructions for usage are referenced in the demucs repository, though not detailed here. The tool is classified under music source separation, specifically targeting audio and music tasks. It is released under the MIT license, allowing for broad use and modification. The evidence does not specify particular user roles or audiences, but the context suggests it is suitable for those working with music source separation models and requiring pretrained weights for such tasks. There is no mention of a graphical user interface, standalone application, or integration with specific platforms; the delivery consists of downloadable model files intended for use within compatible software environments. No information is provided about pricing beyond the open MIT license, and there are no details on deployment by inference providers or specific usage instructions beyond the reference to external documentation. The evidence does not mention any associated services, user support, or commercial offerings.