What do I receive when I purchase a character?
A Character Reference Pack includes 6 production-ready reference assets designed to capture the character’s identity from multiple essential angles
.
5 high-resolution images: front face portrait, side profile portrait, front full-body, side full-body, and rear full-body.
1 high-quality reference video providing additional views of the character’s appearance, proportions, and identity.
Together, these assets give compatible AI image and video tools stronger visual context for creating the character consistently across new scenes, outfits, poses, and productions.
Are Refera characters real people?
No. Refera characters are fictional digital characters and are not intended to represent real people.
All Refera characters are designed as adults aged 18 or older.
How do I use the reference images in an AI video generator?
Most modern AI video generators allow you to upload one or more reference images before generating a video. You can upload Refera character images into compatible tools such as Seedance, Veo, Kling, WAN, and other AI video generators that support image-based generation. The exact interface differs between platforms, but the general workflow is similar: choose the image-to-video or reference-image mode, upload one or more Refera character images, describe the scene, motion, camera movement, clothing, environment, or action you want, adjust any available reference-strength or character-consistency settings, and generate your video. Refera Studios does not guarantee compatibility with every platform or model version, so always check the current capabilities of the AI tool you are using.
Which reference images should I upload for the best character consistency?
For the strongest identity consistency, we recommend using both a front-view image and a side-view or profile image of the character where the AI platform allows multiple reference images. The front view gives the model a clear understanding of the character's main facial structure, while the side view provides additional information about facial proportions, profile shape, nose, jawline, and other identity details. Using multiple complementary views generally gives the AI model more information about the character than relying on a single image. If the platform only allows one reference image, start with the clearest front-facing image.
Why should I use both a front view and a side view?
A single photograph only shows part of a character's appearance. The front view is useful for features such as eye spacing, facial width, mouth shape, overall symmetry, hairline and frontal facial proportions. A side or three-quarter view adds information about the nose shape, chin and jaw structure, forehead profile, facial depth and overall head shape. Providing both views can help an AI model understand the character as a three-dimensional identity rather than simply reproducing one photograph.
Should I upload every image from the reference pack?
Usually, no. More reference images do not automatically produce better results. For many workflows, a small number of clear and complementary references works best. A good starting point is one clear front-facing image and one side-view image. If your AI tool supports additional reference images, you can experiment with more views where they add useful identity information.The ideal number depends on the platform, model, and type of scene you are generating.
How do I change the character's body type?
If you want to change the character's body type, we recommend using only the face reference images rather than full-body references.Then describe the desired body type clearly in your prompt. This gives the AI model a strong facial identity reference without forcing it to closely follow the original body's proportions. For example:
“Use the facial identity from the reference image. Full-body cinematic shot of the same woman with an athletic, strong build, wearing a black running outfit in a modern city at sunrise.”
The exact wording can be adapted to the tool and style you are using.
Can I change the character's clothes?
Yes. You can use the reference images to preserve the character's identity while describing completely different clothing. The original clothing shown in the reference images is not intended to limit how the character can be styled.
How should I prompt the AI model?
A useful prompt usually separates identity from the parts of the scene you want to change. For example:
“Use the same facial identity as the reference images. Medium shot of the character walking through Tokyo at night, wearing a dark tailored suit, cinematic lighting, natural expression, handheld camera movement.”
If you want to change the body type:
“Use the facial identity from the reference image. Full-body shot of the same character with a tall, athletic build, wearing modern sportswear, running along a beach at sunrise.”
If you want to preserve both the face and body:
“Maintain the character's facial identity and overall body proportions from the reference images. Place the character in an elegant hotel lobby wearing a red evening dress, cinematic lighting.”
Different AI platforms interpret prompts differently, so experimentation is part of the workflow.
When should I use full-body references versus face-only references?
Use full-body references when you want to preserve the original build, stature and body proportions. Use face-only references when the scene does not include the characters body, or when you want more freedom to manually alter the body with prompt descriptions.
How do I reduce identity drift across multiple shots in the same sequence?
Identity drift often becomes more noticeable when you ask an AI video model to generate multiple different scenes within a single generation.Each major scene change can force the model to reinterpret the character, which may lead to changes in facial structure, hair, proportions, or other identity details.For better consistency, it is usually preferable to split different scenes into separate generations.
For example, instead of asking for one generation that shows the character walking through a city, entering a car, and then appearing on a beach, create those as three separate generations.This gives the model fewer changes to handle at once and generally makes it easier to maintain the same identity across the final sequence.For more complex productions, Refera reference images can be treated as the master identity source, while each separately generated shot is built around that same identity.
When should I use generated keyframes together with the original Refera references?
The original Refera images should still be treated as the core identity source, even when you are using generated keyframes. Generated keyframes become especially useful once you already have the character placed in the specific styling, wardrobe, environment, or composition needed for a particular shot.
For best consistency, it is often helpful to use both: use the Refera reference images to anchor the character's identity, create a strong still or keyframe in the desired scene, refine that still until the identity and composition are correct, use the keyframe for shot-specific control and keep the original Refera reference image in the workflow as an additional identity anchor where the platform allows it. This is often more reliable than relying on the keyframe alone, because the keyframe helps control the shot while the original Refera reference helps reinforce the character's identity across the generation.
Can I train a LoRA or fine-tune a model using a Refera character?
Yes. You may create a model, fine-tune, LoRA, embedding, or similar asset using the purchased reference images for your own permitted use.You may also use that trained asset to generate content for client projects.
You may not sell, share, distribute, upload, or publicly release a trained model, LoRA, embedding, checkpoint, or similar asset if it allows another person to reproduce or use the Refera character without purchasing their own licence.
Do I own the images and videos I generate?
Refera Studios does not claim ownership of your newly generated output. To the extent that applicable law grants you rights in your generated content, those rights belong to you or the legally entitled rights holder. However, your generated output does not transfer ownership of the underlying Refera character or original reference assets to you.
Can I use Refera characters commercially?
Yes. Commercial use is included unless a product page expressly states otherwise. You may use generated content in:
Advertising
Social media
Films
Videos
Games
Virtual influencers
Client work
Marketing campaigns
Websites
Commercial concepts
And other lawful creative projects. You may monetize the content you create with a licensed Refera character.
Can other people use the same Refera character?
Yes. Refera licences are non-exclusive. Other customers may purchase and use the same character.
Purchasing a Refera character does not give you exclusive rights to that character.
Do I need an account to buy?
No. You may purchase as a guest where guest checkout is available.
You may also create an account. Customer accounts can be used to access order history and previous purchases.
What if I change my mind after downloading?
A change of mind will generally not qualify for a refund once access to the digital product has begun, except where applicable law requires otherwise. For customers with statutory withdrawal rights, the checkout process may require you to consent to immediate delivery of the digital content and acknowledge the consequences for your withdrawal right.
Yes. Commercial use is included unless a product page expressly states otherwise. You may use generated content in:
Advertising
Social media
Films
Videos
Games
Virtual influencers
Client work
Marketing campaigns
Websites
Commercial concepts
And other lawful creative projects. You may monetize the content you create with a licensed Refera character.
