Photanima is an experimental finetune of Anima Base v1.0 to see whether it is a viable architecture for photography. Spoiler alert: it totally is.Turbo LoRA baked in. If you're on a 30-series GPU, I recommend using this with the <a target="_blank" rel="ugc" href="https://github.com/SparknightLLC/ComfyUI-INT8-Toolkit">INT8 Toolkit</a> + INT8 Lazy Torch Compile node for wicked fast gen times. All demo images generated with that combo. These are raw outputs; no upscaling or post-processing.Most demo images contain workflows with custom sigma curve and ODE sampler. These both help significantly with realism. Standalone workflows provided further down this post.❤️ If you enjoy Photanima, you can help offset the cost of training:<a target="_blank" rel="ugc" href="https://ko-fi.com/liftweights">Buy liftweights a Coffee</a><hr /><h2 id="technical-details">🤓 Technical details</h2>v2 is trained on ~2000 images for 45,000 steps. This is an expansion of my <a target="_blank" rel="ugc" href="https://civarchive.com/models/2068592/snakebite-2?modelVersionId=2426502">Snakebite 2.3</a> dataset with around 700 new images and captions reworked for Anima. Training took approximately 48 hours on a Geforce 3090.Pros:<ul><li>Extremely fast.</li><li>Extremely good prompt adherence.</li><li>Anatomy is pretty stable. If it screws something up, changing your steps by +1/-1 usually fixes it.</li><li>Supports up to nearly 2MP with little-to-no distortions.</li></ul>At first, I noticed that Photanima's style was inconsistent - it had a tendency to regress toward a cartoony/CGI look as my prompts became more complex. I was able to mostly overcome this by splitting Photanima into constituent content, style-early, and style-late blocks, then boosted the style blocks well past a strength of 1."Style-late" maps to blocks 7, 8, and 9 - these do alter composition to a degree, so we can't boost them as hard as "style-early."Images are pretty consistent now, but there are some notable drawbacks.Cons in v2:<ul><li>It loses a little knowledge of certain artistic terms like <code>silhouette</code>.</li><li>Microdetail quality is somewhere between SDXL and ZIT. Honestly, it's really good for a 2B model. Two-step upscaling with Anima doesn't help much, but I'm sure the results would be amazing if you sent a Photanima image to a different model for refinement. Or if that's too much work: just add a little film grain. It does wonders and requires no extra VRAM.</li><li>Text capabilities are not as good as those of base Anima. Anything beyond 3 or 4 words is likely going to require numerous re-rolls. This is at least partly due to the Turbo LoRA.</li><li>Excessive fluff tags like <code>masterpiece, absurdres, hyperreal</code> tend to fry the image. The model is photographic and highly aesthetic by default, so there's no need to drive it harder in that direction.</li></ul><hr /><h2 id="recommended-settings-(for-latest-versions)">🛠️ Recommended Settings (for latest versions)</h2>Turbo:<ul><li>6-8 steps. Images often look best at 6, but anatomy is more stable at 8-10, especially with complex prompts.</li><li>er_sde sampler on "ODE" mode.</li><li>Custom sigma curve or simple scheduler: "1, 0.94, 0.9, 0.825, 0.6, 0.5, 0.3, 0.29, 0.2, 0.0"</li><li>CFG exactly 1.</li><li>Preferred resolution: 1040x1520 or 832x1216.</li><li>For maximum realism, begin your prompt with <code>real life photo</code>. If that's not enough, add <code>photo \(medium\)</code> and increase its strength until satisfied. You can usually go up to a crazy strength value like 5 or 6 without breaking the image.</li><li>You can reduce the first number on the sigma curve to 0.95-0.99 to improve realism. This reduces saturation and adds a little noise, but makes the model less stable.</li><li>You can remove NegPip fluff to improve anatomy (e.g. fingers) at the cost of some photographic texture.</li><li>Newest workflow optimized for realism (recommended): <a target="_blank" rel="ugc" href="https://i.ibb.co/JjhHv266/image-photanima-turbo-v2-1.png">Download</a></li><li>Simple workflow with fewer custom nodes: <a target="_blank" rel="ugc" href="https://i.ibb.co/p6t1YcL3/image-photanima-turbo-v2-0-1.png">Download</a></li></ul>Base/Non-Turbo:<ul><li>You can get a good image in 25 steps, but 40 is often better.</li><li>er_sde sampler on "ODE" mode.</li><li>Custom sigma curve or simple scheduler: "1, 0.94, 0.9, 0.825, 0.6, 0.5, 0.3, 0.29, 0.2, 0.0"</li><li>CFG between 3.5 to 4.</li><li>Recommended fluff: "(photo \(medium\):1), real life, score_9, aesthetic"</li><li>Recommended negative prompt: "toon \(style\), anime coloring, painting \(medium\), airbrushed, mutation, distortion, ai-assisted, glossy, shiny, shiny skin, worst quality, score_3, score_4"</li><li>I have found it's helpful to decay conditioning strength from 2 to 1 over the first ~40% of steps. The stock workflow does this.</li><li>Newest workflow optimized for realism: <a target="_blank" rel="ugc" href="https://i.ibb.co/xtrs0F5Q/image-photanima-v2-1b.png">Download</a></li></ul><hr /><h2 id="roadmap">🗺️ Roadmap</h2>I'm pretty excited about the potential of Anima, but let's be clear: I'm not claiming that this checkpoint is a "ZIT killer." The correct model to compare this against is SDXL/IL - and I'm confident that Anima can dethrone it with enough community effort.Directions I'd like to explore next:<ul><li>(✅ Done in v2) There are a handful of Anima "detailer" LoRAs on Civitai. These are not intended for photography, but with enough block pruning, you never know. The right mix could go a long way.</li><li>I suspect further increasing the dataset to ~3k images would help resolve remaining issues related to certain textures or model biases.</li><li>(✅ Done in v2) I'm eagerly awaiting the release of Anima Turbo 1.0. The current Turbo solution is based on Preview3 and I think it's holding back this model's potential a little.</li><li>I'm also looking forward to Anima support in OneTrainer. It will make trying experimental configs a lot less of a hassle compared to kohya-ss. For this v1 run, I stuck with safe values (prodigy, 1.0 LR, no fancy flags.)</li></ul><hr /><span style="background-color:rgb(26, 27, 30);color:rgb(193, 194, 197);font-family:-apple-system, BlinkMacSystemFont, &quot;Segoe UI&quot;, Roboto, Helvetica, Arial, sans-serif, &quot;Apple Color Emoji&quot;, &quot;Segoe UI Emoji&quot;;font-size:16px">Thank you. As always, I look forward to your feedback. Please share the model and upload some images to help it gain traction.

Photanima

v2.0 is trained on over 2000 images.Lots of big improvements:<ul><li>Added over 600 new images to the dataset.</li><li>Specifically selected photos with interesting textures, lighting conditions, or uncommon concepts.</li><li>Names of recurring subjects are now prefixed with <code>@</code>, which should hopefully prevent their features from contaminating general concepts.</li><li>Expanded set of face closeups to reduce the likelihood of "sameface syndrome."</li><li>Switched from Prodigy to ProdigyPlusScheduleFree optimizer, adjusted optimizer parameters.</li><li>Increased support for <code>Asian</code> and <code>African</code> features by reinforcing captions.</li><li>Used custom script for cleaning up issues with JoyCaption outputs.</li><li>Around 15% of the dataset is now captioned exclusively using booru tags; I was aiming for 50% but JoyCaption's tagging abilities are quite a bit worse than its natural language. I will need to test different models (or a higher JoyCaption quant) for tagging in the future.</li><li>Pulled back on use of third-party detail LoRAs.</li><li>Managed to increase strength of Photanima style blocks without damaging composition.</li><li>Upgraded from Anima Turbo LoRA v0.1 to v0.2.</li></ul>

Files with SHA256: 4b0c66a51698e478e4bf9fb87355e1d599fa8fb18891b16f1e71de166a7a9fcf