Short, copyable image recipes that map directly into the AI image studio.
Ultra detailed character design concept sheet on clean white background watercolor and ink wash lean wandering exile faded gray blue cloak frayed tassel edges dark gold tied hair hazel tired vigilant eyes old brass compass on pack faded prayer ribbon on wrist front side back sketches detail callouts small english labels measurement notes arrows dark fantasy costume design dynamic aggressive close up no logo no watermark.
Surreal contemporary artwork inspired by a classical blue green Chinese landscape scroll. Paint mineral pigment mountains rivers pavilions and tiny ancient travelers onto a partially unrolled three dimensional scroll. The landscape rises from the scroll as a delicate 3D relief with layered peaks river valleys and miniature architecture. Clean pure white background, soft studio lighting, refined blue green mineral texture, traditional fine line painting fused with modern 3D rendering, extremely crisp detail, premium art object, no text, no logo, no watermark.
A vast crowd of tiny stylized people in different poses, body shapes, and sizes collectively forms the silhouette of a giant cat. The arrangement is inspired by fractals and emergent patterns. Clusters of people bend stretch and flow to create the cat ears back tail paws and facial features. Wide composition, clean readable silhouette, high detail crowd pattern, elegant graphic surrealism, no text, no logo, no watermark.
On a perfectly pure white background, create a strange and incomprehensible world made entirely from lines of many colors, thicknesses, and curvatures. The world is filled with uncanny biological forms constructed from beautiful mathematical rules such as Fibonacci spirals, parametric curves, and nested geometric rhythms. Wide composition, delicate technical line art, architectural scale, surreal abstract organism design, clean negative space, no text, no logo, no watermark.
A small girl with copper braids and paint stained fingers lies on her stomach across an enormous old library desk in a dusty attic room, chin resting on her folded arms, eyes dreamily half closed as golden afternoon light streams through a round porthole window above her. Towering stacks of illustrated books surround her like a paper fortress, their spines faded to muted sage rust and cream. A half eaten apple and an overturned ink bottle with dried blue rivulets mark her afternoon work. Wide storybook illustration, no text, no logo, no watermark.
Fresh impasto oil paint wall art on a white background. Several large flowers with white and pale pink petals are built from thick sculptural brushstrokes and their centers are soft pale yellow. Add small pink white flowers, green leaves, and white pearl accents scattered around the composition. Bright tactile lively clean decorative art style, strong three dimensional paint texture, no text, no logo, no watermark.
A soft stylized frog resting on a large smooth rock near white sand and a calm pale blue sea. Gentle dreamy light, pastel coastal palette, subtle flowers and grasses near the rock, sculptural storybook texture, peaceful minimal composition, wide frame, no text, no logo, no watermark.
Cute 3D mascot character design of a fluffy baby bunny with a big head, tilted pose, soft rounded body, and a funny innocent expression. High end animated feature character style, delicate fur, warm gentle texture, perfect soft lighting, clean solid color background, polished product quality square render, no text, no logo, no watermark.
Start with a prompt or source image, choose the right model, then keep the output worth using.
Start from the asset you actually have
Use text when the idea is still conceptual, or upload an image when the frame, product, or subject is already right.
Pick the model, ratio, and output settings
Use Veo 3.1 or Seedance for motion, Nano Banana or GPT Image for stills, then set the format.
Generate, compare, and export
Compare the outputs, keep the one with the best composition and finish, then export it for review or publishing.
Compare image and video models in one studio, then choose the right path for motion, edits, or product visuals.
AI Video Maker Models
Use Veo 3.1 or Kling for realistic text-to-video and image-to-video clips, then compare motion before you commit.
Veo 3.1
Kling 3.0
Seedance 2.0
Runway
Seedance 1.5 Pro
Kling 2.6
Wan 2.6
Hailuo 2.3 Pro
AI Image Generator Models
Use Nano Banana or GPT Image for controlled image generation, edits, product shots, thumbnails, and brand visuals.
Nano Banana 2
Nano Banana Pro
Nano Banana
GPT Image 2
Flux Kontext Pro
Flux Kontext Max
Seedream 5 Lite
Ideogram V3
Match the model to the output you need before spending another generation: crisp text, source photo edits, cinematic movement, scene continuity, or fast exploration.
Google introduced Nano Banana Pro as Gemini 3 Pro Image, with stronger reasoning, world knowledge, multilingual text rendering, and up to 4K output. In Zanta AI, it is the image model to try for posters, product mockups, launch visuals, diagrams, and briefs where typography or factual structure matters.
Choose Nano Banana Pro when words are part of the design: poster copy, product labels, diagram text, or campaign layouts that need to stay readable.
The model is built for higher-resolution, polished visual output, so it fits launch pages, ads, decks, and product storytelling inside Zanta.
When a prompt includes layout, factual context, or multiple visual requirements, Nano Banana Pro gives creators a clearer path than a loose mood prompt.
Google describes Nano Banana 2 as Gemini 3.1 Flash Image, combining Flash speed with stronger reasoning, text rendering, subject consistency, and practical output sizes. Zanta AI can use it for campaign concepts, product still drafts, and prompt-led visual exploration.
Nano Banana 2 works well when a creator wants multiple usable options from one brief before choosing a final image direction.
Its text rendering and subject consistency help with product packaging, posters, UI-like compositions, and repeated visual subjects.
The user benefit is format confidence: make square posts, vertical stories, landscape hero images, and wide campaign layouts from the same idea.
OpenAI describes GPT Image 2 as a state-of-the-art image model for generation and editing with text and image inputs. In Zanta AI, it fits close-review stills, brand-safe compositions, realistic details, and transformations that need to follow the prompt carefully.
GPT Image 2 is a strong choice for ads, landing pages, product visuals, thumbnails, and creative mockups that will be inspected closely.
Because it supports text and image inputs, Zanta can use GPT Image 2 for both fresh image generation and source image edits.
When the prompt includes composition, object placement, style boundaries, or edit instructions, present GPT Image 2 as the precision choice.
Google Cloud documents Veo 3.1 for text-to-video and image-to-video, including prompt rewriting, reference assets, first-and-last-frame workflows, C2PA support, and horizontal or vertical outputs. In Zanta AI, it belongs on premium hero loops, product motion, cinematic clips, and source-frame animation.
Veo 3.1 supports prompt-led video and image-to-video workflows, which matches Zanta users starting from text or an approved source frame.
Use this model when realism, camera feel, and visual finish matter enough for a page, ad, or launch surface.
Reference assets and first-and-last-frame workflows make Veo 3.1 useful when the start, end, and subject continuity matter.
Kling VIDEO 3.0 emphasizes multimodal input and output, native audio, multi-shot and storyboard capability, stronger element consistency, multilingual support, and flexible durations. Zanta AI should position Kling 3.0 for character scenes, prop continuity, dialogue cues, and short narrative clips.
Kling 3.0 is useful when the same person, product, object, or setting needs to remain coherent across motion.
Native-audio positioning gives users a reason to test Kling for dialogue-like scenes, sound-aware clips, and expressive video concepts.
This model belongs in prompts with action, characters, props, and multiple visual beats rather than a single ambient drift.
ByteDance Seed describes Seedance 2.0 as a unified multimodal audio-video model with motion stability, reference control, audiovisual generation, and cinematic control over performance, light, shadows, and camera movement. In Zanta AI, it is a natural fit for director-style video prompts.
Use Seedance 2.0 for prompts that describe a push-in, orbit, reveal, pan, handheld shot, or paced product motion.
Motion stability and camera control help users write clearer prompts with subject action, setting, lighting, and lens behavior.
Frame Seedance 2.0 around product reveals, launch scenes, short ad footage, and cinematic social cuts.
Runway documents Gen-4 Video around an input image plus a text prompt, with short clips and common social ratios. Zanta AI can present Runway as a fast first-frame animation option for creators moving from an approved still into motion.
Runway is easiest to explain when the user already has a strong still image: a generated image, product asset, or uploaded campaign frame.
Use it to test a push, drift, reveal, or atmosphere pass before committing to a higher-cost final video model.
Short durations and common ratios make Runway practical for social cuts, creative tests, and lightweight video variations inside Zanta.
Use Zanta AI to draft stills, test motion, and keep campaign-ready assets without splitting workflows.
Use the AI image generator side of Zanta AI to draft beauty, fashion, culture, and creator-led concepts with Nano Banana or GPT Image. When a frame starts to work, move into image to video and test whether the idea has enough motion for shorts, reels, or a campaign teaser.
Turn a landing-page brief into ad hooks, product reveals, and social cuts before the production budget is locked. Use Zanta AI as an AI video maker for text to video concepts, then compare Veo 3.1, Seedance, or Kling when camera feel and the first three seconds matter.
Start from product photos, packaging references, or catalog ideas and make them ready for PDPs, ads, and launch pages. Nano Banana Pro and Flux Kontext fit the AI image editor, enhancer, and upscaler jobs where cleanup, detail, and product consistency matter before a still becomes motion.
Build launch visuals, feature explainers, deck images, and announcement clips from the same creative direction. Use Seedream, Ideogram, or GPT Image for still directions, then use an AI video generator workflow when the story needs a product loop, hero motion, or short explainer.
Practical questions about workflow choice, uploads, credits, and commercial use before you start generating.
Start with text when the scene still has to be invented. Start with an image when the first frame already works. Use Nano Banana or GPT Image for stills, and Veo or Kling when the idea needs motion.
Yes. Zanta can use uploaded source images for video motion or still-image transformations, including portraits, product photography, approved campaign stills, and other brand assets you want to keep recognizable.
The current web workflow accepts common still-image uploads such as PNG, JPG, JPEG, WEBP, and GIF. AI source images support uploads up to 64 MB per file.
Credits depend on the model and settings. Video models like Veo 3.1 and larger image outputs can require more credits than quick still-image tests.
Usually that means the job is scoped too broadly. Narrow the request to one clear scene, one still, or one controlled edit, then retry with a cleaner prompt or a stronger source image. If the first frame matters more than invention, switch from text to an image-based workflow.
Yes. Teams use Zanta for ads, landing pages, product marketing, thumbnails, concept work, and campaign visuals. Final usage still depends on the plan and model terms tied to your account.
No. Zanta runs in the browser, so you can go from prompt or source image to generated output, compare variants, and download the result without opening a separate editor first.
Generate video and image assets in one browser studio, then keep the result that is actually ready to use.