Create consistent portraits, product scenes, and social assets from text or authorized references. Reuse saved models and elements, choose five aspect ratios, and refine selected images without restarting.
1. Select the subject and the right starting mode
From Home, open a saved model to enter its Create workstation. In the left rail, choose Image → Generate image. Use Chat to generate when you want to write the scene yourself; Complete set uses ready-made themed prompts.
Example: For the test, an existing saved model was opened, but the job was an object-only square product scene rather than a portrait.
Likely friction: Generate stays inactive until the selected model loads and the prompt contains text. If you need a recurring fictional subject, create a reusable AI influencer first. See the current Generate Image requirements.
2. Build the scene with text, a model mention, and elements
Describe the subject, setting, composition, lighting, and medium in observable terms. The composer starts with the selected model’s @ mention, which can be removed or added again. Styles and authorized custom elements can be inserted as references; current limits are one element on Fast, four on Ultra 2K, eight on Ultra 4K, and ten on Ultra S 4K. The documented prompt limit is 20,000 characters.
Example: The tested prompt was: An unbranded matte-green serum bottle on a warm limestone vanity, soft morning window light from the left, realistic product photography, label area blank, centered square composition, clean reflections, no people, no text, no logos.
Likely friction: Text alone did not preserve one exact bottle design across the four outputs. For a real SKU, add an authorized product element instead of relying on color and shape words; switching to a lower quality can remove element slots after a warning. Review the prompt-composer controls.
3. Match quality, quantity, and aspect ratio to the job
Choose 9:16 or 3:4 for portrait placements, 16:9 or 4:3 for landscape, and 1:1 for square tiles. Fast returns a batch and does not show Quantity. Ultra 2K, Ultra 4K, and Ultra S 4K expose Quantity choices of one or two on desktop, with the live credit cost updating before submission.
Example: The Nano test selected Fast + 1:1; the workstation showed 20 credits / 4 images before Generate was clicked. Treat that as test evidence, not a fixed price for every account.
Likely friction: Aspect ratio cannot be changed on an existing result. A new shape requires another generation, and choosing two Ultra outputs roughly doubles that run’s cost. Check current ratio and quantity behavior.
4. Generate, inspect, and select a result
Open a result to access Chat to edit, Image to video, Clone Image, Edit clothing, Upscale, Remix, Share, Download, and Details. The available shortcuts can vary between the preview and full detail page.
Example: The first tested Fast result was served as a progressive JPEG at 1024 × 1024. That is evidence for this run only; it does not establish the file format or pixel dimensions of every quality tier.
Likely friction: Nano downloads are watermarked, and upgrading later does not remove watermarks from content created on Nano. Choose the production plan before creating final assets. Read the current download and watermark rules.
5. Refine an almost-correct image with Chat to edit
Choose Chat to edit from the preview. APOB preloads the selected result as the Base image; add a specific Edit description, choose Keep original or another supported ratio, then select Ultra 4K or Ultra S 4K. Generate remains disabled until the required instruction is present.
Example: Change only the wall to warm limestone beige; preserve the bottle silhouette, blank label, crop, counter texture, reflections, and morning light.
Likely friction: A broad edit can alter details that were already approved. State both what must change and what must remain fixed. APOB recommends editing an image that is already roughly 90% correct and regenerating a structurally wrong scene. See the Edit Image guide.
Expert prompts for an AI image generator
For prompts that use a recurring person, first select an original or fully authorized saved model and leave APOB’s prefilled @ mention in the composer. Then paste the relevant prompt below without replacing placeholder text.
Ecommerce seller — product-in-use hero
Purpose: Produce a conversion-oriented vertical scene with space for overlay copy.
Prompt: Demonstrating an unbranded matte-green serum bottle at a limestone vanity, bottle label facing camera, natural hand position, waist-up composition, soft morning window light from the left, realistic skin and product reflections, warm neutral palette, 9:16 framing, clear negative space in the upper-right, no extra bottles, no distorted fingers, no added text or logos.
Expected output: A vertical creator-style product image with a visible item, controlled composition, and copy-safe area.
Social creator — consistent carousel narrator
Purpose: Keep an original fictional character recognizable across a themed post series.
Prompt: An original fictional adult reading-journal host seated at a compact oak desk, annotating a blank notebook, sage-green sweater, soft overcast window light, calm editorial photography, 3:4 portrait crop, preserve the saved model’s facial identity and hairstyle, uncluttered background, no readable book titles, no third-party marks.
Expected output: A portrait that can anchor a carousel while avoiding real-person or protected-title references.
Agency designer — campaign territory board
Purpose: Test a visual direction before spending credits on final assets.
Prompt: Campaign concept for an original meal-kit brand: an unbranded kraft meal box on a cobalt-blue studio counter, sliced citrus and stainless utensils arranged in a disciplined grid, hard side light, crisp shadows, contemporary editorial still life, 16:9 landscape, room on the left for a headline, no people, no existing logos, no text.
Expected output: A landscape concept image suitable for a pitch deck or art-direction review.
Performance marketer — alternate social crop
Purpose: Generate a square ad concept without losing the approved subject and product relationship.
Prompt: Holding one unbranded matte-green serum bottle at shoulder level, direct but natural eye line, pale beige seamless backdrop, softbox key light with subtle rim light, 1:1 crop, product and face fully inside the central safe area, preserve bottle proportions, no duplicate objects, no text.
Expected output: A centered square creative that leaves less risk of automatic placement cropping.
Subject continuity: The
@mention tells the workflow to feature the selected saved model instead of inventing a new anonymous person for every scene.Control beyond text: Written direction can work with predefined styles and authorized custom people, objects, or references.
Visible quality trade-offs: Fast, Ultra 2K, Ultra 4K, and Ultra S 4K separate exploration from final-detail work, while the live Generate control shows the credit cost.
Placement-ready framing: Five ratios cover vertical, portrait, landscape, and square assets before generation, reducing dependence on later cropping.
Connected next actions: Approved images can move into Chat to edit, background removal, upscaling, image-to-video, or talking-avatar creation.
Obtain consent and respect likeness rights
Use a person’s face, voice, or persona only with explicit, verifiable permission for the intended placement. Never impersonate someone or imply an endorsement. APOB’s Terms of Service prohibit unauthorized likeness use and deceptive impersonation.
Confirm commercial rights before publication
APOB’s terms assign generated-content ownership to the user as between the user and APOB, subject to law and third-party rights, while granting APOB a broad service license. Keep releases or licenses for each person, product image, logo, font, and reference used commercially.
Disclose realistic AI-generated content
Label realistic generated or materially altered media with a clear caption or the platform’s AI-content control. Do not strip provenance to present synthetic media as documentary evidence.
TikTok: Label realistic AI-generated image, audio, or video with the in-app setting. TikTok policy
Meta platforms: Use available disclosure controls and retain provenance metadata; qualifying content or ads may receive AI labels. Organic-content labeling and ad transparency
YouTube: Disclose meaningfully altered or synthetic content that appears realistic; qualifying media may receive prominent or automatic labels. YouTube help and 2026 update
Treat copyright as a separate question from platform ownership
Do not request protected characters, imitate a living artist’s identifiable work, or upload copyrighted references without permission. The U.S. Copyright Office’s 2025 report says prompts alone generally do not establish authorship; protectability requires sufficient human-authored expression. Other jurisdictions differ.
Follow every destination platform’s current policy
Disclosure does not excuse misinformation, infringement, unsafe material, deceptive advertising, or prohibited targeting. Recheck policies at publication because rules and labeling interfaces change.
Ecommerce: one product concept, two placements
A direct-to-consumer seller adds an authorized product element, pairs it with a fictional adult model, and approves a square marketplace image before producing a 9:16 social variation. Bottle geometry stays fixed while framing and negative space change.
Social marketing: a consistent fictional creator series
A social team creates a fictional home-organization host, then produces a 3:4 pantry portrait, 1:1 shelf detail, and 9:16 teaser ready for animation. The IAB’s 2025 Creator Economy report projected $37 billion in U.S. creator ad spend for 2025, up 26% year over year, while identifying scale, efficiency, standards, and transparency as active concerns.
Agency concepting: approve direction before full production
An agency tests daylight kitchen, color-block studio, and documentary countertop directions for a fictional meal-kit launch. It presents Fast 16:9 concepts in a pitch deck and renders only the approved route at higher quality.
Object-only publishing: a repeatable still-life system
A solo publisher arranges a blank notebook, pen, and tea cup in 1:1 seasonal still lifes. The AI photo generator changes palette and lighting while preserving composition, without protected covers, logos, or readable third-party titles.
Reusable saved subject. An
@-mentioned model gives recurring portrait and creator campaigns a clearer identity anchor than a fresh text-only description each time.Input control beyond prose. Styles and custom elements can sit beside the written prompt, allowing owned products and reusable references to guide the scene.
Production-aware settings. Four quality tiers, five aspect ratios, and a visible pre-generation credit cost help separate testing from final output.
Connected next actions. Editing, upscaling, background removal, animation, and talking-avatar handoffs reduce asset re-entry between tasks.
The main generation flow is model-centered. A selected model must load before Generate becomes active, which adds setup for someone who only wants a one-off abstract or object-only image.
Higher fidelity consumes more credits. Ultra tiers cost more than Fast, and requesting two outputs on supported tiers approximately doubles the image cost.
Outputs still require review. Hands, small objects, reflections, readable text, and exact product geometry can be wrong; a human must inspect and sometimes regenerate.
Free downloads are watermarked. Nano-account downloads retain their watermark even after a later upgrade, so final commercial production needs plan selection in advance.
No Credit Card Required
Answers about pricing, quality, workflow, privacy, safety, commercial use, and platform compatibility.
What is an AI image generator?
An AI image generator creates or modifies images from text, models, authorized references, or existing images. APOB combines these inputs with styles and reusable custom elements.
Can I use APOB as a free AI image generator online?
Yes. A Nano account can generate images with available credits, and the workstation shows the server-driven cost before submission. Nano downloads are watermarked, and upgrading does not remove watermarks from content created while the account was free, so check the live plan and cost before producing final assets. See the official watermark rules.
What should marketers look for in the best AI image generator?
Assess prompt adherence, reusable references, ratios, resolution, correction controls, visible cost, rights, and privacy—not one gallery example. APOB exposes quality, quantity, ratio, and credit cost before generation, then supports targeted edits.
Can I generate an image from text?
Yes. Chat to generate makes APOB an AI image generator from text. Keep a loaded model’s @ mention when the subject should appear, or remove it for the described scene alone.
Can I use a reference image or image-to-image editing?
Use authorized custom elements to guide a new scene or Edit Image for image-to-image editing. Uploading a likeness, photo, logo, or artwork does not grant usage rights.
How do I keep the same character across images?
Use the same saved model and keep its @ mention in each prompt. Repeat stable identity cues, change only scene variables, and make targeted edits instead of rebuilding an almost-correct image.
Which aspect ratios and quality options are available?
Ratios are 9:16, 3:4, 16:9, 4:3, and 1:1; quality options are Fast, Ultra 2K, Ultra 4K, and Ultra S 4K. Fast returns a batch without a Quantity selector, while the Ultra tiers let desktop users choose one or two images. Confirm ratio before generating because changing it requires a new run. Review ratio and quantity behavior.
Why is my Generate button disabled?
The most common documented reasons are an unloaded selected model or an empty prompt. In edit modes, the required source image or instruction may also be missing.
Can I edit an image after generating it?
Yes. Use Chat to edit, Face swap, Edit clothing, Upscale, or Remove background. A narrow instruction reduces unwanted changes to approved areas.
Can I use generated images commercially?
APOB’s Terms assign Generated Content to you as between you and APOB, subject to the Terms and third-party rights. Obtain permission for any depicted person or protected material.
Are AI-generated images automatically protected by copyright?
Not necessarily. In the United States, prompting alone generally is insufficient. Preserve evidence of human-authored inputs, selection, arrangement, and substantive edits; rules differ elsewhere.
Are my uploaded images and generations private?
The four new results in this Nano test appeared as Private. Sharing a private result requires making it public, and the Help Center says a Nano account cannot switch an already-public item back to private without upgrading. Visibility is separate from data processing: read APOB’s Privacy Policy and avoid sensitive uploads until your organization has confirmed current retention and service-provider terms. See the sharing controls.
What file formats and social platforms are compatible?
The first displayed Fast 1:1 asset in this test was a 1024 × 1024 JPEG. That is a single-run observation, not a promise for every tier or download path; the final Nano download file was not independently captured. Test the extension, dimensions, transparency, watermark, and acceptance by TikTok, Instagram, YouTube, or your commerce platform before producing a batch.
Can I turn a generated image into a video?
Yes. Send a finished still to Image to video or use an authorized face image in Talking avatar. Check hands, text, and product geometry before animation, then follow the destination platform’s disclosure rules.
How long can the prompt be, and how many reference elements can I add?
The documented prompt limit is 20,000 characters. Current element limits are one on Fast, four on Ultra 2K, eight on Ultra 4K, and ten on Ultra S 4K; switching down can remove unsupported elements after a warning. Check the current prompt-composer limits.
