| --- |
| base_model: |
| - ResembleAI/Dramabox |
| pipeline_tag: text-to-speech |
| tags: |
| - diffusers |
| --- |
| |
| UPDATE: |
| I managed to merge in a couple of different LORAs which made a huge difference in it's "abilities". |
| I also released the python script to extract your own Audio weights. |
|
|
|
|
| After some experimentation I managed to extract the Audio DiT Layers of Sulfur and get them working in Dramabox. |
| The result is some added Audio generation "features" - but prompting is a bit tricky and I am still figuring it out :). This is VERY much a WIP. |
|
|
| Usage: |
| This is a drop-in replacement for the dramabox-dit-v1.safetensors file. |
| When using Huggingface for Dramabox, first find the Cache where the model is stored. |
| Once you found that, go to the "blobs" subfolder. |
| There will be one 6.6GB file with a hash as a name. |
| Replace that file with this and make sure to give it the EXACTLY same name. |
| You should now be ready to go! |
| Alternatively you can of course just specify the path to this file instead of the original DiT in your code :) |
|
|
| I am still experimenting with the audio components part as they seem to have modified some things there. |
| Not sure if they even make a difference if I manage to change them aswell. |
