You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Settings
New page in Settings tab, just for Txt/Img to 3D Model
File format for models
Radio buttons or dropdown list for .obj and .ply to turn on only one that user actually needs. Maybe more options in the future, if you'll add support for more formats
Path for saving models \extensions\sd-webui-txt-img-to-3d-model\outputs\ isn't the best option, really, so ability to replace it would be great
Parameters
As already was mentioned here, we need to have Shap-E parameters exposed. Especially seed, otherwise there's no way to get exactly the same result twice.
Must have UI elements Open models directory button. It's way faster than searching through webui subfolders.
Interrupt button to turn off generation with wrong parameters, instead of waiting till it ends.
Now sometimes it's faster to turn off web-ui completely and then restart, than wait until generation with way too big batch size or steps ends.
Things that also might be useful Send to img2img button, that sends screenshot of 3d preview to, well, img2img. Save depth map button, which turns preview into depth map. Would be useful for controlnet and deth2img. Send to img to 3D button in txt2img, img2img, extras tabs
Naming
There's no way to understand which of mesh_1234etc.obj is a gen you're looking for. Would be great to use SD approach: store prompt and seed in the name, when txt-to-3d mode is used.
I'm not sure though, how to make better naming for image-to-3d...
The text was updated successfully, but these errors were encountered:
Thanks @Ninkorr
Some could be implemented easily, such as configuring output format and path, I will do them soon.
Some is still under research such as how to set the seed
But still thanks for the feedback!
Settings
New page in Settings tab, just for Txt/Img to 3D Model
File format for models
Radio buttons or dropdown list for
.obj
and.ply
to turn on only one that user actually needs. Maybe more options in the future, if you'll add support for more formatsPath for saving models
\extensions\sd-webui-txt-img-to-3d-model\outputs\
isn't the best option, really, so ability to replace it would be greatParameters
As already was mentioned here, we need to have Shap-E parameters exposed. Especially seed, otherwise there's no way to get exactly the same result twice.
Must have UI elements
![image](https://private-user-images.githubusercontent.com/136897560/246604972-3e07f73f-59c7-43d3-b9b5-db4e4aa0bf35.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3Mzk4NTk1MzQsIm5iZiI6MTczOTg1OTIzNCwicGF0aCI6Ii8xMzY4OTc1NjAvMjQ2NjA0OTcyLTNlMDdmNzNmLTU5YzctNDNkMy1iOWI1LWRiNGU0YWEwYmYzNS5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjUwMjE4JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI1MDIxOFQwNjEzNTRaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT0zMGQ5NmMyYzZhMDFiMTY1MGY0NWNmMjFhOTE2MDUwODlkZmRjMmUyNDIzOTk1N2YwYTkyNjc4OTBiZDU5YWE4JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCJ9.JY6AvgWuMqKAnqNjhMK50usx3n0KIZeZkat6WQgHcZk)
Open models directory
button. It's way faster than searching through webui subfolders.Interrupt
button to turn off generation with wrong parameters, instead of waiting till it ends.Now sometimes it's faster to turn off web-ui completely and then restart, than wait until generation with way too big batch size or steps ends.
Things that also might be useful
Send to img2img
button, that sends screenshot of 3d preview to, well, img2img.Save depth map
button, which turns preview into depth map. Would be useful for controlnet and deth2img.Send to img to 3D
button in txt2img, img2img, extras tabsNaming
There's no way to understand which of mesh_1234etc.obj is a gen you're looking for. Would be great to use SD approach: store prompt and seed in the name, when txt-to-3d mode is used.
I'm not sure though, how to make better naming for image-to-3d...
The text was updated successfully, but these errors were encountered: