You are able to download all models from the Manager. This should automatically place it into the proper folders, but if you are not using the ComfyUI Manager the creator of InstantID has detailed instructions on their GitHub : github.com/cubiq/ComfyUI_InstantID/tree/main. I hope this helps!
Hey there, thank you for your tutorials, they’re very clear and beginner friendly. A small question if that’s ok, instant ID seems and Face ID seem to work kinda poorly with anime characters as far as consistency goes. Any advice on how to tackle that without training a Lora?
We are so glad you are enjoying the videos! For your question, I need a little more clarity. May I ask are you trying to input a human reference photo and make it in anime style with poses & prompt applied? Or is it an anime reference photo and that you are then trying to generate poses & prompts?
@@compute-grid the latter. I've created an original anime character (I have 30 images now) and Im trying to create multiple poses, expressions, and clothing setups. any ideas for how to approach it? also thank you so much for answering!!!
WOW! 30 images are impressive. Unfortunately though, I don''t believe InstantID will be the best model for you to use as it was trained on human face proportions which is very different compared to Anime faces (Eye size, nose position, etc). It could still work, if the face keypoints line up where they need to be, but you would probably need to find a different model trained on anime faces. If I come across one I will be sure to come back here & let you know.
6 місяців тому+1
Your videos deserve way more views 😁 Liked & Suscribed 👍
I have recreated the entire workflow exactly, using the same settings, the same checkpoint and the same models everywhere - but my result doesn't look nearly as good. Although the pose preview shows the hands correctly, the result is just horribly deformed hands. Every time.
can you add attn nodes for clothing to keep consistency?
Workflow download link?!!
Very helpful. 😊
Is it possible to add character expression changes to this model? Chatacter pose, face and face expression , i think it's perfect.
Excelente tutorial. Espero que sigas subiendo mas videos de Confyui. Gracias
Thanks for keeping these around 10 minutes long. Useful stuff! 👍
Its our pleasure! We are glad you enjoy them 😁
I didn't know about the Face Keypoints. Fantastic. Thanks! ...and HELLO!
Hello! We at AIFuzz really enjoy your channel!!
Thank you so much, that means a lot! ♥️
Is it possible to extract the OpenPose-generated pose data and edit it manually? Then load the manually edited pose back and use it?
Can you please share the work flow
4:35 Hello. Where should we find the model and in which folder should we put it?
You are able to download all models from the Manager. This should automatically place it into the proper folders, but if you are not using the ComfyUI Manager the creator of InstantID has detailed instructions on their GitHub : github.com/cubiq/ComfyUI_InstantID/tree/main.
I hope this helps!
You can use open pose and depth of field for even better results!
Nice!
Thank you to you genius. Now I can see much things by another point of view.
Hey there, thank you for your tutorials, they’re very clear and beginner friendly. A small question if that’s ok, instant ID seems and Face ID seem to work kinda poorly with anime characters as far as consistency goes. Any advice on how to tackle that without training a Lora?
We are so glad you are enjoying the videos!
For your question, I need a little more clarity. May I ask are you trying to input a human reference photo and make it in anime style with poses & prompt applied? Or is it an anime reference photo and that you are then trying to generate poses & prompts?
@@compute-grid the latter. I've created an original anime character (I have 30 images now) and Im trying to create multiple poses, expressions, and clothing setups. any ideas for how to approach it? also thank you so much for answering!!!
WOW! 30 images are impressive. Unfortunately though, I don''t believe InstantID will be the best model for you to use as it was trained on human face proportions which is very different compared to Anime faces (Eye size, nose position, etc). It could still work, if the face keypoints line up where they need to be, but you would probably need to find a different model trained on anime faces.
If I come across one I will be sure to come back here & let you know.
Your videos deserve way more views 😁 Liked & Suscribed 👍
I have recreated the entire workflow exactly, using the same settings, the same checkpoint and the same models everywhere - but my result doesn't look nearly as good. Although the pose preview shows the hands correctly, the result is just horribly deformed hands. Every time.