TMElyralab / MuseV

MuseV: Infinite-length and High Fidelity Virtual Human Video Generation with Visual Conditioned Parallel Denoising
Other
2.43k stars 258 forks source link

Running well, just cannot achieve the effect like your sample vids. #43

Open kacecarol opened 7 months ago

kacecarol commented 7 months ago

I use the default settings MuseVRun,load an image and got really deformed outcome... Can you please release a guide to explain what the parameters do and how to generate a decent video, thank you.

itechmusic commented 6 months ago

@kacecarol hi, can you upload your input and output? we can update guidance later.

kacecarol commented 6 months ago

They are all default. prompt: (masterpiece, best quality, highres:1),(1girl, solo:1),(beautiful face, soft skin, costume:1),(eye blinks:1.8),(head wave:1.3) img_edge_ratio:1 video_len:12 seed:851 randomize wid:563/height:714 videostep:10 first_redraw_step:30 first_redraw_guidance_scale:3.5 model:majic uselcm:false

03409-1240738044-(((masterpiece))),((best quality)),_lora_add_detail_0 7_,_lora_GoodHands-beta2_1_ ,full body,body shot,1girl,slender,solo,extrem

https://github.com/TMElyralab/MuseV/assets/81342264/2d662101-176b-4cb0-b713-b327b2a1f911

kacecarol commented 6 months ago

Interesting I downloaded an integrated pack of MuseV from bili (BV1ux4y1v7pF) and it worked well.Not as comfyui.

phighting commented 6 months ago

Interesting I downloaded an integrated pack of MuseV from bili (BV1ux4y1v7pF) and it worked well.Not as comfyui.

@kacecarol Great to hear that! Could you share your results? Curious to see..

kacecarol commented 6 months ago

Interesting I downloaded an integrated pack of MuseV from bili (BV1ux4y1v7pF) and it worked well.Not as comfyui.

@kacecarol Great to hear that! Could you share your results? Curious to see..

https://github.com/TMElyralab/MuseV/assets/81342264/64abceef-cbf4-481d-a929-f4cd25fca3f0

far from perfect, but make sense.