What Actually Goes Into a Baking Template
A baking template is just a saved renderer configuration that pre-sets all your shadow, GI, and diffuse settings so you can apply them across multiple objects without reconfiguring everything each time. It saves you from manually setting ray trace depth, shadow map resolution, and filter size every single time you set up a new asset. Most people find themselves spending more time configuring these settings than they should, and a template cuts that down significantly. I work with Cycles mostly, and I built a template years ago that handles a lot of the repetitive setup. Once you get it working right, you just attach it to whatever geometry you're baking and hit the button. The whole thing goes from maybe 20 minutes of setup per project to just a few clicks.
Setting Up a Baking Template in Blender
Open your file, select your object, and go to the Render Properties tab. Under the Cycles section, set your light cache samples to something reasonable like 4096 for final renders, and bump your diffuse bounces up to at least three if you're doing interior shots. Set shadow map resolution to 2048 minimum unless you're working on something that requires significantly higher detail. Once those settings look right, go to the Node Editor, switch to the Shading workspace, and create a new Texture node. Set the type to External or Non-Color depending on whether you're baking maps or regular textures. Then in the Output Properties, set your bake type — Direct Only, Indirect Only, or Combined depending on what you need. Name your template file something that makes sense later, like "Interior_Bake_Template_v3". Save it to your user presets folder so you can access it from any file. After that, you can go to File > Templates > Render and assign your saved config. When you open a new project, it loads automatically with all your settings already configured. This alone saves me probably 15 minutes per project just on setup time.
Common Issues People Run Into
The biggest problem I see is people not understanding that a baking template doesn't fix bad UV unwraps. If your UV islands are overlapping or your texel density varies wildly across your mesh, the baked result will look terrible no matter how good your template settings are. Always check your unwrap before you bake anything. You can do this by creating a simple checkerboard texture and applying it to your object in Material Preview mode. Another issue is that baked results won't match between different render passes unless you keep your template consistent. I had a project where two artists were working on the same scene and one was using a template with 512 sample shadows while the other used 1024. The lighting didn't match during compositing and I spent six hours fixing it. Make sure everyone on your team is using the same template version. Light map resolution scaling is also a thing that catches people off guard. If you bake at 1024x1024 for a small prop and then try to scale that same map for a large wall, you'll get blurry results. The workaround I use is to bake at a base resolution and then use a denoiser afterward rather than just cranking up the sample count, which slows everything down without necessarily improving visual quality.
Get the Full Details

When a Baking Template Won't Help
Baked templates assume your scene lighting is static. If you need dynamic lighting that changes in real time — think day/night cycles, moving light sources, or interactiveGI — a baking template is the wrong approach. You'll need a fully real-time lighting setup instead, which means investing more in your hardware or going with lighter materials that don't require extensive ray tracing. Also, baked textures take up memory. A single high-resolution baked map for a detailed character can easily be 4K or 8K, and if you're baking multiple maps per object you're looking at significant VRAM usage. This matters if you're working on a console or mobile project where memory budgets are tight. In those cases, consider using a mix of baked maps and real-time lighting rather than going fully baked. If your scene has a lot of translucency or subsurface scattering, baking becomes problematic. The approximation in a baked map will never fully capture how light passes through skin, wax, or thin fabrics. You'll get muddy results, especially in areas where the light needs to travel through the material. For those materials, real-time SSS is the better option if your pipeline supports it, even though it costs more performance.
Using Your Baking Template Efficiently
Once your template is saved, the main workflow is straightforward. Open your scene, apply the template, set your output path, choose your bake type, and run. For most organic models with moderate detail, this takes between two and five minutes on a decent GPU. A full scene with multiple characters and props might take ten to twenty minutes depending on complexity. The thing nobody tells you is that you should always do a test bake at low resolution first. I learned this the hard way on a project where I baked an entire scene at full quality only to realize my UV layout had major issues. Had I tested at a lower resolution first, I would have caught that in five minutes instead of waiting forty minutes for a failed bake. Set your test resolution to 1024x1024 and check for anomalies before committing to the full bake. You also want to keep your template versions organized. I name mine with dates and major changes noted, like "Interior_Bake_2024-08-15_GI_Improved". The rendering industry moves fast and settings that worked six months ago might not be optimal now. Having a versioned system means you can go back and compare results if something looks off in a later build.
Finally, document what each setting in your template does. I keep a simple text file alongside my template that explains why I chose each value. When someone new joins the team or when you come back to an old project, it saves a lot of confusion about why certain values are set the way they are. A lot of people skip this step and then spend time later trying to reverse-engineer their own settings from months ago.
