FYI . . . The link doesn't download the completed workflow. The Canny ControlNet portion isn't connected, the math nodes aren't connected or are connected improperly and the guide has nothing in it about adding the switch for switching between T2I and I2I. The Empty Latent in the Samplers Group doesn't appear to be connected to anything. Should it be connected up to the switch or am I missing something? It's probably me missing something, LOL!
Also, you could add the CR SDXL Aspect Ratio Node (from the Comfyroll Sudios Pack) to remove three nodes in the Generation Inputs Group (Height, Width, Batch Size) and 1 Node from the Samplers Group (Empty Latent). This would reduce the overall number of Nodes and simplify the workflow a bit.
Thanks for the great work! I appreciate it and built my workflow through your tutorial and am now trying to simplify and condense it.
I worked through all four stages of this tutorial yesterday and it was really excellent building it up piece by piece, projects likes this are such a better way to learn. Thank you for all the hard work that went into it. Parts 3 and 4 felt a little more rushed in terms of explanation, they took a lot more double checking of the text and images than the earlier parts did. Nonetheless, it made for an excellent practical learning experience.
The only part I really struggled with was the "That means we will have to schedule 40 steps (25/0.75=40) and skip the first 15 steps." section. So I think I must have missed something because by my maths 25/0.75 is 33.33 and so my Total Scheduled Steps was 33, Skipped Steps was 8 and Base Sampler End was 28. Where did I go wrong?
FYI Rgthree has a very good context switcher that works very efficiently in comfy, alongside a useful seed node that solves most newcomer's problem : the second iteration following the rendering, losing the initial seed, especially in batches. Evidently sequential solves this and the seed can be recovered from pasting the png in a new workflow, but i found his solution elegant.
FYI . . . The link doesn't download the completed workflow. The Canny ControlNet portion isn't connected, the math nodes aren't connected or are connected improperly and the guide has nothing in it about adding the switch for switching between T2I and I2I. The Empty Latent in the Samplers Group doesn't appear to be connected to anything. Should it be connected up to the switch or am I missing something? It's probably me missing something, LOL!
Also, you could add the CR SDXL Aspect Ratio Node (from the Comfyroll Sudios Pack) to remove three nodes in the Generation Inputs Group (Height, Width, Batch Size) and 1 Node from the Samplers Group (Empty Latent). This would reduce the overall number of Nodes and simplify the workflow a bit.
Thanks for the great work! I appreciate it and built my workflow through your tutorial and am now trying to simplify and condense it.
I worked through all four stages of this tutorial yesterday and it was really excellent building it up piece by piece, projects likes this are such a better way to learn. Thank you for all the hard work that went into it. Parts 3 and 4 felt a little more rushed in terms of explanation, they took a lot more double checking of the text and images than the earlier parts did. Nonetheless, it made for an excellent practical learning experience.
The only part I really struggled with was the "That means we will have to schedule 40 steps (25/0.75=40) and skip the first 15 steps." section. So I think I must have missed something because by my maths 25/0.75 is 33.33 and so my Total Scheduled Steps was 33, Skipped Steps was 8 and Base Sampler End was 28. Where did I go wrong?
FYI Rgthree has a very good context switcher that works very efficiently in comfy, alongside a useful seed node that solves most newcomer's problem : the second iteration following the rendering, losing the initial seed, especially in batches. Evidently sequential solves this and the seed can be recovered from pasting the png in a new workflow, but i found his solution elegant.
Any idea on how to install comfyui manager if we're using stableswarm on google colab?
Honestly, I have no idea, let me see if someone from the team can answer this one