⚡ Anima Workflow:
🛠️ Purpose & Design Philosophy
This workflow is designed for quality and autonomy, not speed. It follows an "all-in-one" philosophy: configure your settings, hit queue, and let the workflow handle everything from initial generation to high-res detailing in a single pass.
Not for Speed: If you want rapid-fire generations, this is not the tool for you. A solid, much faster alternative created by darksidewalker can be found here.
Personal Use: This was built for my personal production needs. It is not intended to be a "one-size-fits-all" solution, but I am sharing it for those who value the same high-fidelity results. Please adjust the settings to your preferences!
Heavy Duty: Due to the multi-stage processing, this workflow can be resource-intensive. In my experience, the detailers are not usually needed. YMMV.
On v1g and after: If the results are too blurry for you after USDU, you can try using the RTX nodes after it to help.
v2 is compatible with AIO versions of Anima models. Note: You might have to select a random model in the Checkpoint Loader node even if you are not using it. This is because it references your models/checkpoints folder. The opposite may also be the case if you are using the Checkpoint Loader node but not the Model Loader in the Diffusion Model Loader group which references the models/diffusion_models folder.
🚀 Key Features
Beyond standard generation and upscaling, this workflow integrates:
Power LoRA Loader: Efficiently manage multiple LoRAs without spaghetti wires.
Global Controls: Centralized Seed, Sampler, and Scheduler nodes for a unified experience.
Bypass Control to toggle features on/off.
Visual Validation: Integrated Image Comparer nodes to see exactly how your image evolves at every stage.
Upscaling: 2-stage upscaling using standard image upscaling and Ultimate SD Upscale (optional).
Triple Detailer Groups: 3-stage detailing using standard BBOX and SEGM detection models for faces, hands, and clothes.
CivitAI Ready: Images are saved with full metadata (Model, LoRAs, Prompts) for easy site parsing.
⚠️ Disclaimer & Compatibility
Install at Your Own Risk: Updating ComfyUI or adding custom nodes can break your environment. I am not responsible for any installation issues.
Portable Version: This was built and tested on the ComfyUI Portable version. Desktop app users may require additional troubleshooting.
"Your Mileage May Vary": Your environment is almost certainly different from mine. I do not guarantee 1:1 compatibility.
Nodes 2.0: I strongly recommend disabling Nodes 2.0. It causes unpredictable behavior; I will not provide support for any issues arising from its use.
🤝 Support & Boundaries
No DMs: Direct messages are disabled due to high volume. Please use the Discussions tab below. Check previous comments first (unless there aren't any yet), as most common questions may have already been answered.
Custom Requests: I do not take private requests for custom workflows. If you need a specific solution built, please post a Bounty on CivitAI. There are many talented creators here who will be happy to assist you for a fee.
Modifications: You are free to add or remove nodes as you see fit. However, if you change the internal logic, you are responsible for your own troubleshooting.
The only place I am actively maintaining this workflow is here on civitai. If my workflows are being posted and monetized elsewhere, whoever posted them is obligated to provide support.
Description
Release Notes: Anima Workflow v1d
This update introduces flexibility for prompt generation and refines the core node structure to ensure maximum stability. Anima v1d focuses on integrating natural language processing power while maintaining a clean, functional workspace.
New Features & Logic Updates
1. Impact Wildcard Processor
To take full advantage of Anima’s natural language capabilities, the Impact Wildcard Processor has been integrated.
Connectivity: This node is linked to the second text input for the positive prompt.
Why it’s here: There is a vast selection of high-quality wildcards available on CivitAI that pair perfectly with the Anima model's understanding of descriptive language.
Flexibility: If you prefer manual prompting, this node can be bypassed or deleted without affecting the rest of the logic.
2. Node Reliability (Subgraph Unpacking)
Based on stability testing, the subgraph containing the Global Seed and Global Sampler has been unpacked.
These components now reside as two separate nodes on the main canvas.
This change ensures the seed incrementation and sampler logic function exactly as intended, avoiding the processing errors found in nested configurations.
Performance Benchmarks
This workflow is optimized for efficiency while maintaining high detail. Under default settings, a full generation cycle completes in approximately 242 seconds on an RTX 5060 Ti (16GB).
The full cycle includes:
Initial Image Generation
Latent/Pixel Upscaling
Ultimate SD Upscale (USDU)
Two dedicated Detailer passes
Note: Performance may vary based on your specific CUDA environment and hardware. Please feel free to adjust steps and resolution to fit your specific needs.
FAQ
Comments (6)
I will be dropping Anime Enhancer from the workflow on the 1e update. My experience with it has been mixed and I prefer consistent results. It's a nice tool, but not for me. I will be replacing it with Lying Sigma Sampler from the Detail Daemon custom nodes. This node is basically a simplified version (only 3 settings) of Detail Daemon and is giving me consistent results.
My experience with the Anima Enhancer has been mixed in regards to the replay feature as well. It feels more like a stabilizer than a detailer but I don't personally find the reduced saturation and detail roulette worthwhile over just rolling the dice again. With that said, it might be improving artist styles (just noticed this today and need to test further).
I believe Detail Daemon is "incompatible" with Anima in the sense that, unlike with SDXL, messing with the sigmas seems to drastically shift the color profile of generations. Even with the detail amount set to 0, blues start becoming green to the point where sky photos can flip to forest photos unprompted. It also tends to darken images, which people may or may not want. I've found myself mostly avoiding it with Anima except for turning light-colored styles on their head for the lulz.
I did some quick tests with Lying Sigma Sampler and I find that it has a similar effect on the colors but setting the value low enough seems to keep the color change within a more tolerable range while adding details. The values you have work pretty nicely so people shouldn't get huge color shifts as long as they don't mess with the dishonesty factor too much.
Rescale CFG (0.70) from your Illustrious workflow has a similar coloring effect for some reason and also tends to drastically affect the composition, but the added detail can be nice. I do think it's a bit too random for a general Anima workflow. I might start using Lying Sigma Sampler instead. Nice find there!
Here's a quick album for those interested. I think you've pretty much worked out an ideal setup as far as these options go. It's kind of wild how differently these nodes work in Anima compared to SDXL models.
(wasted half a day last week figuring out why all my gens had weird colors until I noticed my minimized Detail Daemon node was enabled...)
And this unrelated, but using an upscale model paired with USDU can cumulatively cause pretty drastic color shifts which can be addressed with a color matching node. The Color Corrector node kinda tends to do its own thing even with a reference image. KJNodes and Easy Use both have color match nodes (or there's this new one which I have yet to test since I'm heading out on vacation tomorrow). It isn't really a fire-and-forget option when the sampling might improve the colors, so this probably isn't something you'd add to the workflow. Just figured I'd mention it while I'm bombing you with a wall of text anyway haha
@mac2492 no worries on the length of the message and thanks for the input!
I lean more toward liking darker compositions when generating images. I am a long time user of Better Days and its previous iterations.
With Anima, I keep Lying Sigma Sampler's detail_amount at -0.01 to -0.02 usually. According to Gemini these values would be similar to 0.10 to 0.20 on the normal Detail Daemon Sampler node. I also switched to using this on my currently unreleased Illustrious workflow and also will be dropping RescaleCFG.
I have used other color correction nodes before, but prefer the Color Corrector. The Color Match node form KJ nodes has given me some issues before (if I recall it errors out when bypassed or if bypassed it send the referenced image though instead of the target image). Maybe they fixed it since then, but it's an easy pass for me because of that.
That new color match node sounds interesting, but I think the caveat for me is that it sounds like it needs a reference image to function, which leads me to think it will have the same 'unable to bypass' issue that I had with the Color Match node previously.
I think if folks want to use something else for color correction, it's not a hard thing for them to just replace the current color corrector node with whatever they prefer.
@Gladas I totally missed that -0.01 was roughly equivalent to 0.10 and not 0.01. I've only tested a little but I can see why you'd swap!
For the color matching, I actually have mine in the USDU section while still keeping the Color Corrector node at the end for final adjustments. The logical location would either be between the model upscale and USDU, or after USDU depending on your philosophy. The upscale models themselves change the colors in different ways, like RealESRGAN Anime 6B tends to saturate while Remacri tends to desaturate. I've mostly swapped to RTX Video Super Resolution since it's super fast and seems match/beat UltraSharpV2 without really touching the colors (not a suggestion but it is pretty nice xD). Anyway, USDU then does IMG2IMG on the upscaled image with color drift which can cause even more color drift before the image is thrown at the detailers which can again cause color drift. For some images, I found that the colors were impossible to salvage by the time it reached the Color Corrector node.
I actually haven't been color matching as often (though I generate a preview just in case) simply because the upscale method I'm using is pretty good about color, but with a typical Remacri -> USDU setup I found that a lot of my colors were simply too far from the original. Here's an album for a more visual demonstration (helps to open the images in separate tabs). The colors aren't so much bad as they are simply different, and this just me using the first image that was spit out rather than an extreme case. The bright pink ribbon is dulled and approaching violet while the distinct indigoish splash in the hair is almost lost. I'm not very technical with color but sometimes the colors just hit right and the drift was killing me slowly. I'm sure this is splitting hairs for many people so it's less a suggestion and more food for thought!
@mac2492 nice, that upscaler node you mentioned sounds like an easy swap in for myself. I try to avoid using Upscale Models since they seem to overdo it (for me), so this is right up my alley. I'll probably keep the node out of the workflow when uploading since I can't know for certain that everyone who uses my workflow is using Nvidia hardware. If the colors are good enough on the final result when using this, then I (or others) can just bypass the Color Corrector. Thanks for sharing the album to see the comparisons! I'll have to try it myself and compare.
@mac2492 I went ahead an uploaded a version using that upscale node and a couple other RTX nodes. Works like a charm!