Style Grading

Preserve your vision on framing, posing, lighting and blocking. With Style Grading remain the Director of your creative work.

Beyond the functional aspect of visual storytelling, storyboards are your vibe conveyors. They're your mood boards, your visual pitch, encapsulating the artistic direction of your project. With Style Grading for key shots, you're not just planning; you're inspiring. You're giving your team, your clients, or even yourself, a glimpse into the soul of your work.

Describe the scene

The description (or prompt) is critical, without it the AI is just working off colored pixels. Words give context. Your prompt should be clear, and note all the important elements.  Some tips from here.

  • Tag what you see, not what you know. Don't use tags of stuff that is not visible in the final image. If the character has jeans but the framing is at upper body, don't mention it.
  • Be explicit. Avoid abstract concepts and prefer visual ones. For example, "writing" is better than "doing homework".
  • Minimal Tagging Criteria. Too much tags cause noise and lead to less precise results. Keep things to the strict minimum, put down only what you absolutely want to see.
  • Use common words. Use simple, common and straight-forward keywords. You would typically prefer "excited" than "ecstatic". Keep in mind this does not mean using less specific words ("shirt") over more specific ones ("blouse"), that would still be recognized.

 If you want your characters to look consistent consider always refer to them specifically and consistently. A powerful trick is to say something like “Man who looks like <famous person’s> brother.” The man will consistently look similar but not exact— and it’s sorta how a lot of casting starts tbh. There’s a lot of guesswork with prompts, so try a bunch. It’s best to remember perfect is the enemy of the good— especially in previs. Don't cut hung up on perfection, keep moving forward.

Pick your Style

Tap the Style button (initially, “Comic”) on the lower left to change or adjust the Style settings.

There are only two settings.

‍Style menu, Comic, Photo, Clay etc. The last ones in the list give you access to  the latest generic Stability Diffusion (or potentially other) image generators, without any specific "style", so it will react only to your prompts.

Strength slider. The higher this value the more “creative” the Style Grade will be.  A higher strength will usually result in a stronger effect, but at the cost of ignoring your original composition — so people and objects may move or vanish/appear. Play with it to get a feel.

Generate the Style Grade

On the main Style Grade screen, Tap the large Magic Wand button in the lower right.  Magic! *

* magic is tricky. You may not like the result! If so, you can tap the wand to try again. You can adjust the strength or prompt and then try again. You have 10 tries per shot, per session.

What’s the catch?!

There are some difficulties with the current setup, it can be hard when multiple people are in the frame for the prompt to know which actor to apply what words to.  It can also be difficult to achieve consistent looking actors between shots.  Both of these were actively looking to improve in the next few months, and there are some tricks to address them below.  But in general, it’s best to think of Style Grading as good for the equivalent of key frames or cover shots; not for every shot in a storyboard.  That will let you get your vision across without bogging you down trying to get every single shot perfect.

Tips & Tricks

  • Beards, etc. If you want to add something specific, like sunglasses or a mustache you can say “wearing sunglasses” or “Mustache” and often that does the trick. Using "Touch Up" afterwards is usually simpler for this, but obviously the closer you get initially the better.
  • Clothes, etc. This works a bit for clothes too, within a close approximation of the shape of what they’re currently wearing.  Take the same scene and try adding  “1950s” or “Ancient Rome” or “Klingons” to the prompt and see how it changes their hair and clothes. You may need to up the Strength for a good effect. Again, Touch Up is better.
  • No Beards, etc. The Negative Prompt is very useful for removing errors in a generation. For example if the first try adds a beard to a character and you want to remove it, try adding the word “beard” to the Negative Prompt and see what happens.
  • Consistent actors. Make your model as close to your vision for your actor as possible. Give each actor a specific colored shirt. ReadyPlayerMe actors tend to have more detail which can be helpful for the AI. Also as noted above, try to refer to them in the prompt as looking like someone specific and famous, or their brother/sister.  “Man in a red shirt who looks like Obama” etc.  But it will be impossible to get 100% consistent anything with Style Grading, it’s more about vibe or mood, and key shots to share.
  • Any Image Fast. If you set Strength to Max, it will totally ignore your image. This can be useful for generating images quickly instead of trying to source an external image for use in your storyboard.
  • Annoying Nanny Filter. If you hit an error 400, it means the image generation decided to censor you. We don’t like this either but it’s not something we have control over.  A good way to to avoid is to use euphemisms.  For example, if you ask it to make grown up fully clothed couple kiss, it may stop you. But if you instead say “lock lips”… (silly, but works.)

Advanced Settings

In general if you want to learn about the advanced settings we can recommend this article published at this link.

But if you do, you should probably export your shots as clean jpgs and then head over to places like Stability.ai or Midjourney, or OpenAI and dig into their tools directly.

Our goal is to provide something at the intersection of quality and simplicity for you to nail your pre-visualization work without slowing you down.  This is a trade off to keep us focused on the job preproduction!

A Note on Privacy

Unlike the rest of the Previs Pro which has no “cloud” or internet aspect by design, and therefore no security surface to leak data, to create the Style Grading effect we need to interface with an online Image Generation service hosted by stability.ai. You can read about their privacy policy here.

https://stability.ai/use-policy

https://stability.ai/privacy-policy

We only share a flat image of your shot, the prompt, and some technical settings.

Beta

Each time an image is generated it costs us a small amount of money which is why there are limits and we label this feature “beta” until we see how it’s going.  There’s lots of potential cool stuff we can add to Style Grading, so stay tuned.

‍

‍

‍