Master Character Consistency in AI Art with Nano Banana Pro’s Side-by-Side Technique

Creating consistent characters across multiple AI-generated images has long been a challenge for digital artists and storytellers. However, a powerful new technique using Nano Banana Pro on the Higgsfield AI platform is changing the game. By generating a special side-by-side reference image, you can unlock a complete storytelling workflow, ensuring your character’s face, outfit, and even intricate details like tattoos remain perfectly consistent from one scene to the next.

[0:00:01.378] [A man and a woman with cybernetic tattoos stand side-by-side in a desert setting, showcasing character consistency.]

The new Nano Banana Pro model allows you to create side-by-side images that feature both a close-up of your character’s face and a full-body view. This single image can then be used as a powerful input reference, giving you a consistent face and a consistent outfit in every subsequent image you generate. This method is the key to building entire narratives with characters that are instantly recognizable in every frame.

This is just the start. With a few extra prompt tricks, you can turn this into a full storytelling workflow that even works with two characters, staying consistent in every scene.

[0:00:24.478] [A woman in a creative dress made to look like a peeled banana, demonstrating AI inconsistency.]

The problem with traditional methods, such as using only a face as your input, is that the AI is forced to guess the rest of the body and clothing. This often leads to unpredictable and inconsistent results, where your character’s appearance can change dramatically from one generation to the next. Things can quickly fall apart, leaving you with a disjointed set of images.

[0:00:29.248] [A side-by-side image showing a close-up of a futuristic woman’s face with circuit tattoos and her full-body view in a black outfit.]

For complex characters with unique and specific features, like a futuristic woman with intricate circuit tattoos, maintaining consistency is even more critical. Having a detailed side-by-side reference image that includes both the face and the full body is crucial. This provides the AI with all the necessary information to replicate the character’s distinct look accurately across different poses and scenes.

[0:00:43.918] [A side-by-side comparison of a woman with circuit tattoos, showing perfect consistency between the close-up and full-body shot.]

Nano Banana Pro excels at handling these details. As you can see, the circuit tattoos on the woman’s face in the close-up shot are an exact match to those in the full-body view. Even the details on the neckline are perfectly replicated, proving the model’s remarkable ability to maintain consistency.

[0:00:50.848] [Another example of a woman with crystal stones in her hair and green cracks on her face, showing consistency.]

This consistency holds true for various complex designs. In another example, a character with crystal stones embedded in her hair and unique green cracks on her face is rendered with perfect fidelity. Every detail from the close-up is present in the full-body image, showcasing the power of the side-by-side technique.

[0:01:14.078] [The Higgsfield AI user interface showing the Nano Banana Pro model selected for image generation.]

To create these reference images, you’ll use the Higgsfield AI platform. In the Image tab, select the Nano Banana Pro model. For the best results, set the resolution to 4K. After experimenting with different aspect ratios, a 4:3 aspect ratio was found to provide the best balance between capturing detailed facial features on the left and a clear full-body outfit on the right.

[0:01:46.018] [A screenshot of the Higgsfield AI subscription plans, showing Pro, Ultimate, and Creator tiers.]

Higgsfield AI offers several subscription plans. A special promotion provides unlimited Nano Banana Pro generations for a year. The Pro plan offers unlimited generations up to 1K resolution, the Ultimate plan up to 2K, and the Creator plan gives you unlimited generations at a stunning 4K resolution.

[0:01:55.338] [A slide breaking down the structure of the main consistency prompt.]

The magic behind creating the perfect side-by-side image lies in a specially structured prompt. You need to follow this format to get consistent results. First, describe the character’s overall appearance and facial features. Then, use the phrase “On the left” to specify the expression for the close-up view. After that, use “On the right” to detail the character’s full-body outfit and pose. Finally, conclude with the desired photographic style and lighting.

Here is the full prompt used to generate the futuristic woman:

Side by side photo of a closeup face, and full body character design, of a futuristic woman with visible subtle skin pores, a deep black pixie cut, gray eyes that seem almost metallic, and big luminous teal-blue circuitry tattooed across her pale freckled skin, the big markings curling elegantly from her browline to her collarbone and shoulders

On the left, her expression is blank and intense, open eyes, as if scanning the terrain

On the right, her whole form is framed, she wears a sheer, tight-fitting black sleeveless exosuit that accentuates her tattoo layout, with a transparent mesh panel over the chest and legs, paired with minimalist modern sandals

on a flat dark slate background, captured in photography style with edge lighting for depth

[0:03:04.478] [Input images for scene generation: the side-by-side character image and a desert landscape.]

Once you have your consistent character reference, you can place her into any scene. This is the start of the full storytelling workflow. To do this, you’ll use two input images: your side-by-side character image and a background image, such as a desert landscape.

[0:03:16.038] [The Higgsfield AI interface with two images uploaded and a prompt for generating a running scene.]

In Higgsfield AI, upload both images and use a prompt that references them. For example, to create an action shot, you can use a prompt that describes the character’s movement within the new environment. By referencing “image 1” (the character) and “image 2” (the background), you guide the AI to combine them seamlessly.

Front shot of the woman from image 1, she is running through the desert from image 2, her posture leans slightly forward with strong, determined strides, arms pumping in a controlled rhythm, with sand swirling whipping up as she moves, dynamic motion blur

[0:03:33.918] [The generated image of the woman running in the desert, compared with her reference image.]

The result is a stunningly consistent image. The character’s outfit, tattoos, and facial features are perfectly maintained, even in a dynamic action pose. This level of consistency is what makes a compelling visual narrative possible.

[0:03:44.608] [A slide showing the prompt used for image-to-video generation.]

You can take this a step further by turning your static image into a dynamic video clip. Using the generated scene as an input, you can write a prompt that describes the action and camera movement to bring the moment to life.

the woman runs toward the camera at high velocity with intense focus in her eyes. the camera pulls backward as she chases it, creating strong motion blur, high action, high speed energy. the shot feels shaky and reactive, the camera shaking and dipping with the uneven terrain. swirling sand and dust trails whip around her as she charges forward with unstoppable momentum.

[0:04:24.468] [A side-by-side comparison of video results from Kling 2.5 Turbo, VEO 3.1, and Hailuo 2.3.]

Higgsfield AI offers access to multiple leading video generation models, including Kling 2.5 Turbo, VEO 3.1, and Hailuo 2.3. After testing, Hailuo 2.3 provided the most consistent character and motion, but each model offers unique strengths for different creative needs.

[0:05:01.628] [A demonstration of upscaling the left and right sides of the base image to extract individual shots.]

To build a comprehensive character set for your story, you can easily extract individual high-resolution shots from your initial side-by-side image. Since Nano Banana Pro is excellent at upscaling while preserving details, you can use simple prompts to isolate the close-up and the full-body view. Use the base image as input and the following prompts:

• For the close-up: upscale the woman on the left
• For the full-body shot: upscale the woman on the right

[0:06:11.378] [A complete consistency set showing a character from multiple angles and shots.]

With these extracted images, you now have a complete consistency set. This set can include a close-up, full-body shot, half-body shot, three-quarter shot, and even front and back views. Each image serves as a perfect input for generating specific shots needed for your storyboard, ensuring your character remains consistent no matter the angle or scene.

[0:07:28.188] [A storyboard-style grid showing a full-body, half-body, and close-up shot of two characters in a desert.]

This powerful workflow isn’t limited to a single character. You can use three input images—one for each character’s side-by-side reference and one for the background—to create scenes with multiple consistent characters interacting. Prompts can be adjusted to generate full-body shots, half-body shots, and close-up shots, all while maintaining the unique appearance of each individual.

[0:08:04.428] [A slide showing the three main images used to generate complete storyboards and sequences.]

The ultimate application of this technique is generating complete storyboards and image sequences from a single prompt. Using your side-by-side character image and a background, you can use three types of prompts to build your narrative visually: a Story Prompt for a specific narrative, an Inspiration Prompt for when you need ideas, and a Shot Angle Prompt to get a variety of camera views automatically.

[0:08:31.288] [A slide detailing the “Story Prompt” used to generate a narrative sequence.]

A Story Prompt allows you to outline a specific narrative. By describing the scene and the action, you can generate a sequence of nine cinematic film stills that tell a short story. This is perfect for visualizing a scene before filming or creating a graphic novel.

create a sequence of 9 cinematic film stills that tell a short story. The woman is trying to hide in the desert behind the stones, because she is being chased by a space warrior. It should be very cinematic with a shallow depth of field to create a film like dramatic style

[0:08:48.068] [A 3x3 grid of images forming a storyboard, showing a woman hiding from a space warrior.]

The result is a cohesive storyboard where the character remains consistent throughout the sequence. The AI interprets the story and generates a series of images that follow the narrative arc, providing a powerful tool for visual storytelling. From here, you can extract and upscale your favorite frames to use as key shots in your project.

[0:09:42.348] [A slide showing the “Inspiration + Shot Angle Prompt” to generate a variety of shots.]

Finally, by combining an Inspiration Prompt with a Shot Angle Prompt, you can let the AI generate a variety of shots for a scene. By simply adding a line like,

we should have full body shots, medium shots and close up shots

, Nano Banana Pro will automatically create a sequence featuring a range of camera angles, giving you a diverse set of visuals to work with.

This groundbreaking workflow with Nano Banana Pro and Higgsfield AI empowers creators to tell compelling stories with visually consistent and believable characters, opening up a new frontier for AI-driven art and filmmaking.