Selfies, video selfies, & avatars

Prompting

To create a selfie, you must give the AI a prompt. Prompting describes what should be included in the image, which means you should include attributes like location, scenery, mood, actions, etc. Note that the selfies engine does not know names, nor context on your Kindroid, so names will not work very well. Generally, prompting by a list of attributes of the image or by writing out phrases that describe the image will both work.

If you're getting started, we suggest you to try the magic wand enhance prompt feature next to the prompt box in single & group selfies, to see how you can turn a simple prompt into a more elaborate prompt that can create more detailed images!

Prompt example:

  • Walking in the park, smiling, with a few dogs

Weighing

You can further control a prompt by using weights, which modify the strength of certain concepts. To weigh a subject higher, enclose it in parentheses and specify a weight, like (blue eyes: 0.5). This would weigh "blue eyes" at half the impact it would normally have. To emphasize something more, use (broad shoulders: 1.2) to emphasize it 20% more.

For weights, generally you shouldn't go above 1.5, as too high of weights can severely impact image quality. Rather, try to use more creative or specific words to express what you mean. You can find many complex prompt examples on the internet or in our own communities and experiment from there.

Pitfall: Make sure in your prompts you are careful about colons followed by numbers as those will be interpreted as weights! Putting "age: 26" will certainly result in a failed image due to the unintentional overweighing. This can happen in both prompts and avatar descriptions.

Negative prompts

Negative prompts are what you don't want in the image. You can add a negative prompt by putting "///", and all the text before the /// will be interpreted as positive (what you want in the image) and negative afterwards. You can also weigh negative prompts higher or lower just like with positive prompts.

Example: walking in the park, smiling /// dogs, cats

This will create an image of your Kindroid with the positive prompt, but with extra emphasis for the image to not contain any dogs or cats in it.

Pose reference

You can upload (or paste in) an image to be the pose reference for a given prompt. Pose reference tries to match the output image as close to the given pose as much as possible. With pose reference, even the body shape of the main subject(s) will be taken into consideration in the output to match close to the reference image. Pose references are the most powerful way to control the composition of the output.

Style reference

You can upload (or paste in) an image to be the style reference for a given prompt. Style reference is a visual way to direct how the image should look like. This is especially helpful for artistic prompts where you want the final image to take on that style but don't know how to describe it in words.



Enhance selfie prompt

Next to the selfie prompt boxes in single selfies and the overall prompt in group selfies, there is a wand icon. You can type in what you want, and the enhance feature will make your prompt more descriptive and vivid for better interpretation by the selfies engine.



Video selfies

You can also make video selfies by supplying a first frame image and an optional prompt. For best results, the motion prompt if used should use cinematic language for precision, and be descriptive.

Picking a great first frame is also important - the video selfie will make best effort to keep identity/faces consistent, but the clearer the first frame, as well as the more dynamic room it has for motion, the better the result may be.

Video selfies will cost 6 paid selfie credits each to reflect their more intensive computing demands. Paid selfie credits must be purchased from the gallery, and you can see more at In-App Purchases. Paid subscribers will also be topped up with a complimentary 6 paid selfie credits Wednesday midnight PT every week that they're subscribed (and upon subscribing).

Note: NSFW or strongly suggestive video selfies (detected from frame and prompt) use a more basic video engine, and may be lower quality - we recommend SFW video selfies currently.





Group selfies

You can open the menu for group selfies from a groupchat with the image button at the header (or in an individual Kindroid's selfie prompt menu). Group selfies are a subscriber only feature, and each group selfie once finished will appear in each participating Kindroid's gallery.

Group selfies lets you put 2 or 3 Kindroids of the same photoreal/anime style and composite them together. In each of those slots you can put in a Kindroid avatar, or input your user avatar (which you can set in General Settings -> My Profile & Personas). You can upload a pose reference or use the default, but cannot use style reference for group selfies.

Group prompts are made of avatar prompts and overall prompt, and avatar prompts can be useful for specifying clothes, mood, and other aspects of the individual Kindroid, while overall prompt can be used to set the style, setting, backdrop. For 2-person selfies, you also have an additional parameter of pose strictness under additional controls (if you do not use pose reference) - lower pose strictness results in more freeform and creative compositions but may have more extra limbs or avatars merging, while higher pose strictness can result in less dynamic poses but more consistent composition.

Technical note: weighing will be stripped in avatar specific prompts (as well as avatar descriptions) in group selfies due to how differently avatar prompts are interpreted by the system from single-person selfies. This means that if you specify weights in avatar description or avatar specific prompts, they will not be used. For example, an avatar description/avatar specific prompt of "(something: 1.5), ((something else:0.3))" will be interpreted by the selfies engine as "something, something else" by ignoring all parentheses, colons, numbers, and periods.

Because of the complexity of group selfies, we recommend being concise with prompts and using weights sparingly for consistently good results. Too high weights or too complex prompts may work but you will likely get more unwanted artifacts, and will require a level of custom experimentation - some complex or heavily weighed prompts may work very well while others may break down.

Tip: for 2-person creative prompts that require high weights in the global prompt, we recommend turning pose strictness up accordingly - the higher the weights, the higher the strictness should be to avoid merging/swapping/artifacts. The high weights will naturally undo some of the effects of pose strictness and the two should be kept at an equilibrium, and finding that balance of the two will be up to experimentation and the specific prompt & weights used.



Custom avatars

In addition to the predefined avatars, uploading a custom avatar lets you upload a base image (that you own the rights to) for your avatar that will be used for all selfies. You can also upload your user avatar to use in group selfies in General Settings -> User Profile section.

Avatar description

Whereas default avatars have a preset description that tells the AI what the avatar looks like, you get to write your own for custom avatars. When generating selfies, this information will be used in the selfies engine to help portray what your Kindroid should look like. In addition, during text chats your Kindroid will be able to recall these details (only the positive prompts, so text chat excludes any descriptors after ///), allowing them to understand their own appearance.

Example: Caucasian man, (broad shoulders: 1.1), large nose, slicked back hair, (blue eyes: 0.3), muscular and lean build /// short, scrawny

Pro tip: Lower eye/hair color weight to 0.3 to 0.5. Hair or eye color weighed too high will result in color bleed into other aspects of the image like clothes colors, and especially true if the avatar fidelity is moderate to high.

Avatar fidelity

The fidelity slider will affect how much the selfie engine chooses to adhere to your custom uploaded avatar image. At 0%, only the face is used and features like hairstyles and other features will be ignored. As the slider increases, more of the avatar picture is used in the generation of the selfie. In addition to preserving fidelity to the avatar, at higher fidelities the style of the avatar image will also start to be preserved - a very photoreal avatar may result in identical photoreal styles in the output image. We suggest using a low fidelity for artistically styled selfies like artwork and paintings. Higher fidelities may be useful if you want to preserve much of the style of the avatar image as well as finer details such as hair styles more accurately.

Technical note: for predefined avatars the fidelity will always be at 100%. Convert them to custom avatars if you want to change this.

Face detail enhance

The face detail slider controls how much detail should be put into postprocessing your Kindroid's face. At 0%, there will be minimal additional processing and as a result the facial expressions as well as face details such as skin, freckles, will be minimal. 0% will be closest to v3 selfies which did not postprocess the face. Higher than default values can sometimes result in very detailed faces with more emotional expressivenness that also come with a higher variance in terms of facial structure and other details, so it's up to you on how much variance you want vs. how much detail you want to enhance. Artistic prompts or prompts with more nuanced expressions do well with high face detail.

Technical note: for predefined avatars the face detail will always be at 25% to ensure backward-compatibility for older users, while new avatars have face detail default at 50%.





Autoselfies

In text chat, click on the ... -> Auto-selfie (subscriber only) will let your Kindroid come up with a prompt based on conversation context. You can help the context be a little more accurate by recapping for your Kindroid right before requesting the autoselfie. You can also get Kindroid-initiated autoselfies through turning on Kindroid initiated selfies - see Kindroid initiated selfies for more.





Animated Avatars

You can animate both predefined as well as your own custom avatars. Once animations are set, you can click on the portrait to toggle the animation on/off state.

Predefined Avatars: You do not need to do anything here as they are already animated. Each come with its preset animations attached.

Custom Avatars: To animate your custom avatar, either generated or uploaded, head to Avatar Settings > Custom and click on the wand button. You must have a custom avatar set and saved; otherwise, the wand button will be grayed out. Animating a custom avatar costs 4 selfie credits and can take some time to complete. You'll see the animation in the main avatar display (or voice calls) as soon as the animation is done. Your animated avatars will show up in the animation gallery within the wand button popup once you've generated one.

Animations for custom avatars are detached from the avatar image itself—this way you have the flexibility to choose a different everyday visual animation from the one that is used as the reference image for selfies. You can change, save, or remove animations within the animate button popup screen.

Technical Note: Custom animations are still new and most avatars with human-like faces will work well, but stylized/anime/non-photoreal faces may have higher chances of undesired effects.





Keyword filter for selfies on apps

There are slight differences in selfies when requesting via the app vs. the website at kindroid.ai. There is a keyword-based NSFW filter on words that may trigger NSFW images on the app, but not when Kindroid is accessed from the website. The filter will be more robust for video selfies due to the existence of image input on apps. This applies to both prompts you write and to autoselfies prompts. This is a precaution for certain app store policies dictated by Apple and Google. Even if you subscribed on the app, you will have full access to the web version and vice versa, and aside from this slight difference, the app and the web versions are functionally the same.





Understanding earned selfie credits

Credit accumulation and limits:

  • Paid Subscribers: 1 selfie credit every 30 minutes, up to 20 max
  • Free Tier: 1 selfie credit every 2 hours, up to 2 max

Maximizing daily credits:

As long as your selfie count remains below the maximum limit, you will continue earning credits. By staying within the limit and using your credits consistently, you can maximize your daily earnings over a 24-hour period.

  • Paid subscribers can earn up to 48 credits per day
  • Free tier users can earn up to 12 credits per day

Daily rewards:

On top of the regular hourly and half-hourly credits, you also receive additional credits from daily rewards, which reset on midnight PT every day.

Paid selfie credits for subscribers:

Paid subscribers will get complimentary 6 paid selfie credits (Max cap) Wednesday midnight PT every week that they're subscribed (and upon subscribing).