Frequently asked questions

You get 3 free credits when you sign up, enough to generate around 1-3 images depending on which model you pick. No credit card required. After that, credit packs start at $4.99 for 50 credits and credits never expire. There is no subscription. You only pay when you run out and want more.

#is-renza-free

Click any Generators category from the home page (Avatar, Logo, Wallpaper, etc.) or go straight to /create. Pick a style, type a short prompt describing what you want, and hit Generate. The first result is ready in 3-15 seconds depending on the model.

#how-do-i-start

For the free client-side tools (resize, convert, compress, crop, collage maker, HEIC to JPG/PNG) you don't need an account, everything runs in your browser. For AI generation, video, editing, and enhancement you need to sign in so we can track your credit balance. Sign-in is via Google or a magic link sent to your email, no password.

#do-i-need-an-account

Depends on the model. Flux Dev (the default) is 6-12 seconds. Hyper Realistic and Ideogram are 8-20 seconds. Nano Banana 2 is 10-25 seconds. Videos take longer, 1 to 5 minutes depending on the model and duration.

#how-fast-is-it

Every generation costs credits depending on the model. Flux Dev and Hyper Realistic cost 2 credits. Ideogram and Nano Banana 2 cost 3. Edit models (Flux 2 Max, Nano Banana Pro, GPT Image 2, Qwen Image Max) cost 2-4. Video models cost 5-25 credits depending on length and quality. Credits are deducted only on a successful generation, refunded automatically if the model fails.

#how-do-credits-work

No. Credits never expire. Buy a pack, use them over weeks, months, or years. There is no monthly minimum and no subscription. Your balance stays until you spend it.

#do-credits-expire

The Starter pack is $4.99 for 50 credits, that's around $0.10 per credit. The Standard pack at $14.99 gives 200 credits (effectively $0.075 per credit). The Pro pack at $39.99 gives 600 credits ($0.067 per credit). Bigger packs = better unit price. See /pricing for the current full list.

#pricing-packs

No. renza is pay-as-you-go only. You buy credits when you need them, they never expire, and you decide when to top up. We may add a subscription tier later for very heavy users, but the default model stays one-time top-ups.

#subscription

Yes. If you bought credits and are unhappy, email info@nicomelian.com within 14 days of purchase and we refund unused credits. If a specific generation failed and credits weren't auto-refunded by the system, contact us, we refund failed generations no questions asked.

#refund

All major credit cards (Visa, Mastercard, American Express, Discover), debit cards, Apple Pay, Google Pay, and SEPA bank transfer for European customers. Payments are processed by Polar, a regulated Merchant of Record, they handle the card details, we never see them.

#payment-methods

Some AI models are dramatically more expensive to run than others. Flux Dev uses 28 inference steps; premium models like Nano Banana 2, Nano Banana Pro and GPT Image 2 are proprietary with higher per-image API costs we pay. We pass the cost on proportionally so you pay only for the quality you choose. The credit cost per model is shown next to every model picker in the app.

#why-different-credits

For image generation: Flux Dev (default), Hyper Realistic, Ideogram, and Nano Banana 2. For image editing: Flux 2 Max, Nano Banana Pro, GPT Image 2, and Qwen Image Max. For video: Veo, Kling, Seedance, and LTX. For enhancement: Topaz upscaling.

#which-models

Hyper Realistic is our pick for photoreal portraits and product shots — strong on skin, texture and lighting. Nano Banana 2 is excellent for both portraits and compositing but costs more credits. Flux Dev is a solid all-rounder for general use.

#best-model-for-photoreal

Nano Banana 2 handles anime/illustration best. For aggressive stylistic transforms (photo to anime), use Nano Banana via our /photo-to-anime tool. For pure anime generation from text, Nano Banana 2 with a clear style prompt gives the best results.

#best-model-for-anime

Ideogram is purpose-built for legible text in images, posters, social cards, mockups. Nano Banana 2 and GPT Image 2 also handle text well. Most other models produce gibberish on letters longer than 3-4 characters, so reach for Ideogram when text matters.

#best-model-for-text

Flux Dev. It's a strong middle-ground choice: high quality (28 inference steps), supports reference images, costs 2 credits. For premium quality or compositing your trained model into a scene, switch to Nano Banana 2.

#default-model

Be specific. 'a dog' is bad; 'a fluffy golden retriever puppy sitting in tall grass at sunset, soft warm light, shallow depth of field' is good. Three rules: describe the subject in detail, describe the lighting and mood, describe the camera or render technique. Length matters, 30-80 words is the sweet spot. Avoid contradictions ('photo of a cartoon' confuses the model).

#how-to-write-prompts

24 built-in styles: minimalist, cyberpunk, watercolor, anime, pixel art, 3D render, oil painting, sketch, vaporwave, photorealistic, cartoon, gothic, flat, fantasy, isometric, low poly, art deco, pop art, Studio Ghibli, noir, pastel, graffiti, baroque, and futuristic. Each maps to a specific prompt modifier that biases the model toward that aesthetic.

#how-many-styles

12 categories: logo, tattoo, avatar, wallpaper, sticker, YouTube thumbnail, product photo, illustration, character, pattern, landscape, and food. Each comes with its own aspect ratio default and a tuned prompt template that produces results appropriate to the category.

#how-many-categories

Square (1:1), widescreen (16:9), vertical (9:16), classic photo (4:3 and 3:4). Each category has a sensible default, wallpapers default to 9:16, YouTube thumbnails to 16:9, logos to 1:1. You can override per generation from the studio.

#aspect-ratio

Yes, for the models that support it (Flux Dev and Nano Banana 2). The reference biases the generation toward the composition, colors, or subject of the reference. Useful for matching a brand style, replicating a pose, or iterating on an existing image. Drop the reference in the studio and the supported models switch to image-to-image mode automatically.

#reference-image

Yes. Hit Generate again with the same prompt, image models are non-deterministic, so each click gives a different result. To bias toward variation, tweak one adjective. To bias toward consistency, copy the seed (shown in the result panel) and reuse it.

#variations

Go to /edit, drop your photo, pick a model (Nano Banana Pro is a strong default), type what you want changed in plain English (e.g. 'replace the background with a beach at sunset' or 'make the sweater navy blue'), and click Generate. The AI keeps your subject identity intact and changes only what you describe.

#how-to-edit

Nano Banana Pro is a great all-purpose default. Nano Banana is best for subtle, identity-preserving edits. GPT Image 2 is most aggressive for stylistic transforms. Qwen Image Edit excels at preserving fine details. Each model card on /edit lists what it does best.

#edit-models

Yes, use /tools/object-remover. Describe what to remove ('the person on the left', 'the lamp post', 'the watermark in the corner') and the AI fills the area with whatever should naturally be there. No manual masking needed.

#remove-objects

Use /tools/background-changer. Upload the photo, describe the new background ('modern studio with grey backdrop', 'tropical beach at sunset', 'professional office'), and the AI replaces the background while keeping your subject lit naturally for the new scene.

#change-background

Yes, /tools/photo-restoration. Scan the photo at high DPI if you can, drop it in, and the AI removes scratches, recovers fading, sharpens detail, and keeps the original era and identity. For severely damaged photos, run it twice, each pass cleans more.

#restore-old-photos

Yes, /tools/colorize-photo uses Qwen Image Edit tuned for naturalistic colorization. Add the year to your prompt ('photo from 1942') for era-appropriate clothing and skin tones. Restore damage first (/tools/photo-restoration) if the original is also damaged.

#colorize-bw

Four: Veo 3.1 (Google, top tier with audio), Kling 2.5 (cinematic, supports start + end frame), Seedance 2.0 (native audio, multi-shot), and LTX Video (cheapest, fastest). You can pick per-generation from the /video studio. Costs range from 5 to 25 credits depending on model and length.

#video-models

Typically 5-10 seconds per generation. Some models support up to 20 seconds. Longer videos cost more credits. For most use cases (social media, intros, B-roll), 5-8 seconds is enough.

#video-length

Yes. From the /video studio, drop a start frame and the AI animates from it. You can also add a prompt describing the motion you want. This is the most reliable way to get a specific look, generate a still image first (or upload your own), then animate it.

#video-from-image

Some models support audio generation (Veo 3 is the strongest). For the others, output is silent. You can mute or add your own track later in any video editor.

#video-audio

Use /photo-to-anime. Drop your photo, the prompt is pre-filled with the right style (cel-shaded, expressive eyes, Studio Trigger / Makoto Shinkai feel), and Nano Banana redraws your photo as anime art while keeping you recognizable. Takes 5-15 seconds.

#photo-to-anime

Yes, /photo-to-ghibli uses GPT Image 2 (the same model behind the viral ChatGPT Ghibli trend) to convert photos into Miyazaki-style art. Landscapes work especially well; portraits also preserve identity. Output is high-res and free of any watermarks.

#photo-to-ghibli

Use /photo-to-pixar. We use GPT Image 2 to render you as a 3D Pixar-style character with subsurface scattering on skin, large expressive eyes, and the warm cinematic lighting from Inside Out or Up. Bright daylight selfies convert cleanest.

#photo-to-pixar

Yes, /photo-to-painting uses Nano Banana Pro. Output has real brushstroke texture, classical chiaroscuro lighting, and museum-quality detail. Common use: wedding or family photo as a printed canvas. Output is high-res enough for canvas prints up to 16x20.

#photo-to-painting

Yes, /photo-to-cyberpunk relights your face with magenta and cyan neon, adds wet streets and atmospheric fog, and gives it the Blade Runner 2049 cinematic feel. Works on any selfie. Identity is preserved, just the lighting and environment change.

#cyberpunk-photo

No. The seven client-side tools, image resizer, converter, compressor, cropper, HEIC to JPG, HEIC to PNG, and collage maker, run entirely in your browser. No upload, no signup, no credits used. Your files stay on your device.

#tools-need-account

Yes. /tools/heic-to-jpg converts HEIC files to JPEG. /tools/heic-to-png does the same for PNG (lossless). Multiple files at once. Conversion runs in your browser via a JS decoder, your photos never leave your device.

#heic-conversion

/tools/image-compressor. Drop the image, drag the quality slider, and watch the file size drop in real time. Quality 80-85 typically saves 70-90% of file size with no visible difference at normal viewing sizes. WebP output saves another ~30% over JPEG at the same quality.

#compress-image

Yes, /tools/image-resizer. Enter the exact width or height you need, lock aspect ratio toggle keeps proportions, and download the resized file. Common use cases: email attachments, social media profile photos, web upload limits.

#resize-image

/tools/collage-maker. Pick a grid layout (2-up, 3-up, 2×2, 3×2, 2×3), drop a photo into each cell, tweak the gap and background color, and download a single combined image. Output is up to 3000×3000, high enough for printing.

#collage

Depends on your account type. Free accounts (no purchases): new generations default to Public, they appear in our explore feed at a /g/ URL and may be indexed by search engines. Paid accounts (any purchase): new generations default to Private, visible only to you. You can toggle the visibility of any individual generation from /account at any time.

#are-my-generations-public

No. We do not train models on your prompts, your uploaded photos, or your generations. The AI models we use (Flux, Nano Banana, GPT Image, etc.) belong to their providers and we don't supply them with training data either.

#do-you-train-on-my-data

Email info@nicomelian.com from the address on your account. We delete your account, all your generations, and all associated data within 30 days. Payment records are kept for 7 years as required by tax law, but those contain no personal photos or prompts.

#delete-account

On Cloudflare R2 (S3-compatible object storage) encrypted at rest. Private generations have unique non-guessable URLs and require your session to view. Public generations are served from the same CDN at predictable /g/ paths.

#where-are-images-stored

Yes. Everything you generate on renza is yours to use for any purpose, including commercial work, client deliverables, merchandise, printing, resale, advertising, you name it. We don't watermark, claim ownership, or restrict usage beyond a few sane defaults (no depicting real people without their permission, respect existing trademarks).

#commercial-use

You own the output. Renza grants you a perpetual, royalty-free, worldwide license to use, modify, and redistribute every image and video you generate. The underlying models (Flux, Nano Banana, etc.) also do not impose royalty obligations on output created via the renza pipeline.

#license

Only if they've consented or if the person is a public figure being depicted in a clearly non-defamatory way. Generating intimate imagery, defamatory imagery, or impersonations is against our terms and may result in account termination. The AI does not refuse outright but you are responsible for how you use the output.

#real-people

Yes. Many users print renza-generated art on canvas, posters, T-shirts, and merch and sell on Etsy, Redbubble, Society6, Printful, etc. We don't take a cut. The only restriction is content that violates third-party trademarks (no fake Nike logos, no Pokémon characters, etc.).

#sell-prints

JPEG, PNG, WebP, GIF, AVIF, HEIC, and HEIF. Maximum file size is 15 MB per upload. Output is delivered as WebP (smallest), JPEG (most compatible), or PNG (lossless with alpha), you pick per generation.

#image-formats

Most image models output at 1024×1024 (square) or equivalent for other aspect ratios. Premium models can hit 2048×2048. For larger output, generate at native and then run through /tools/image-upscaler (powered by Topaz) at 2× or 4× to reach print-ready sizes.

#max-resolution

All modern browsers: Chrome, Safari, Firefox, Edge, Brave, Opera. On mobile: iOS Safari and Android Chrome. The free client-side tools (resize, convert, etc.) use the Canvas API and work in every browser since 2018.

#browser-support

Not publicly. We run on top of fal.ai for most models, if you need direct API access, fal is open and accepts new accounts immediately. For automated batch workflows on top of renza specifically (with our pricing, our presets, our credit system), contact us and we'll set you up.

#api-access

Two options: Google (one-click) or a magic link sent to your email (no password). The magic link expires after 5 minutes. We don't use traditional username+password to reduce phishing risk.

#sign-in-options

Email info@nicomelian.com from your current address with the new address you want. We migrate the account manually within 24 hours.

#change-email

Polar (our payment processor) emails a receipt automatically after every purchase. For VAT/tax invoices with company details, email info@nicomelian.com with your purchase confirmation number and the company info you want on the invoice. We respond within 2 business days.

#invoice

Polar (our processor) does not tell us exactly why a card was declined, it could be insufficient funds, an issuer block on cross-border transactions, or 3D Secure failure. Try a different card, or use Apple Pay / Google Pay if your bank app supports them. For European customers, SEPA bank transfer is also an option.

#card-declined

Three common causes: (1) Wrong model for the job, try Nano Banana 2 or Hyper Realistic for higher fidelity. (2) Prompt is too short or contradicts itself, be more specific. (3) For styles, Nano Banana 2 handles anime/illustration better than the Flux models. Each model card lists what it's best at.

#image-quality-bad

Check three things: (1) Your credit balance, if zero, the request is rejected and you'll see an error. (2) Your prompt, extremely short or empty prompts can fail. (3) The selected model status, occasionally a fal provider has an outage. If credits were deducted but no image arrived, refresh your account page, the system auto-refunds failed generations within a minute.

#no-image-generated

Three fixes: (1) Use Nano Banana Pro or Nano Banana, they preserve identity better than aggressive models like GPT Image. (2) Lower the strength setting (in the editor). (3) For headshot generation, upload a clear daylight selfie shot from the front, the AI has more identity signal to lock onto.

#face-identity

Check spam first, magic links often land there. The link expires after 5 minutes; if you waited longer, request a new one. If still nothing, try the Google sign-in option instead. If Google also fails, email info@nicomelian.com and we'll help.

#cant-login

Polar webhooks usually deliver credits within 5-10 seconds. If after 2 minutes credits still don't show, refresh /account. If still missing, email info@nicomelian.com with your Polar order ID, we credit manually within an hour.

#credits-not-added