This is slower and more resource intensive than most TTS models so keep that in mind. But I do get more expressive results.
You can voice clone or create a custom voice but I find voice cloning works much better.
I had an issue with auto downloading
I downloaded from https://huggingface.co/ResembleAI/Dramabox/tree/main
within /models make a folder called drama box
download
dramabox-audio-components.safetensors
And from the assets folder download
silence_latent_frame.pt
This is slower and more resource intensive than most TTS models so keep that in mind. But I do get more expressive results.
Description
FAQ
Comments (3)
what if the output has added speech or there is way too much silence before or after the speech? What setting should I try changing first?
Ive been doing prompts like I would do in LTX, for me what works best is like the man says ,"I really like boats!" He gets angry and says, "why did I buy a boat!" almost a caption for each one or two sentences. Estimating isnt bad, you could leave it at 0 or adjust if needed.
The json has "FranckyB/ComfyUI-DramaBox" in it. It'll fail out of the box. Still can't get the Options node loaded even after fixing the json.