Image FX

What is Image FX?

Image FX is a browser-based generative media workspace built around a simple premise: most people do not need one AI model, they need access to several, and they do not want to manage a separate subscription for each. The platform positions itself as an all-in-one creative workspace for AI-powered image and video generation, bringing several current-generation models together in one interface. In practice that means a marketer, a small-business owner or a freelance designer can open a single tab, write a prompt, pick whichever engine suits the job, and walk away with a finished asset — a product shot, a stylised illustration, a sharpened archival photo or a short animated clip.

The product line splits into three connected areas. The first is image creation from text, where a prompt becomes a still image. The second is image editing and enhancement, covering style rewriting, upscaling, background work and inpainting on material you already own. The third is motion, where a still frame or a written description becomes a short MP4. All three draw on the same credit balance, which is the structural decision that shapes almost everything else about how the tool feels to use.

One clarification matters before going further, because the naming collision in this category is genuinely confusing. Image FX here refers to the commercial service at image-fx.app, operated under the name Image-Fx AI and Image-Fx Labs. It is not the Google Labs tool of a similar name that lives on the labs.google domain, and the two products share nothing beyond a naming collision. The confusion is compounded by the fact that this platform does offer Imagen 4 as one of its selectable engines — but offering a model is not the same as being the model's publisher, and reviews of the Google tool describe a different product with different access rules and a different pricing story entirely.

Core Features

• Text-to-image across multiple engines: The headline capability turns a written description into a finished image. Instead of paying for several separate services, you pick a model per job inside the same workspace, because each model excels at different styles. The selectable roster spans Imagen 4, Seedream, Qwen Image, GPT Image 2 and Nano Banana Pro alongside the platform's own ImageFX Pro engine, with resolutions running from 1K up to 4K.
• Per-model credit pricing: Rather than a flat per-image fee, cost tracks the underlying compute. The house engine sits at 4 credits, Seedream at 6, Qwen at 8, Imagen 4 at 10 and Nano Banana Pro at 20. That spread is the mechanism that makes experimentation affordable: you draft cheaply and spend on the final render.
• Style transfer and image-to-image rewriting: Upload an existing picture and rewrite it — into an oil painting, a sketch, an anime treatment or a different colour story — while keeping the underlying composition. A strength slider from 0 to 100% governs how far the model departs from your original, which is the difference between a subtle grade and a full reimagining.
• AI upscaling to large formats: A dedicated enhancer enlarges images by analysing edges, texture and detail rather than simply interpolating pixels. Output is capped at 20,000 pixels on the longest side, which is generous enough for large-format print work as well as web delivery.
• Image-to-video and text-to-video: Still frames become short clips with described motion — a slow push in, a 360-degree product rotation, a drifting camera. Text-to-video skips the source image entirely and builds the clip from a written scene description, with the option to attach up to two reference images to hold visual consistency.
• A supporting toolkit of utilities: Around the four main workspaces sits a set of narrower tools — background remover, inpainting, outpainting, image resizer, watermark remover and an image-to-prompt reverse describer — that handle the small fixes which otherwise send you back to a desktop editor.

Use Cases

1. E-commerce product imagery on a short timeline: A seller with a plain catalogue photo can rewrite the background into a styled lifestyle scene, upscale the result for a marketplace listing that demands high resolution, then animate the same frame into a rotating product clip for a paid social placement. What used to require a photographer, a retoucher and a motion designer becomes an afternoon of iteration by one person.
2. Social content in multiple aspect ratios: Campaign work rarely needs one image; it needs the same idea in a feed square, a vertical story and a wide banner. Because the video tools expose 16:9, 9:16, 1:1, 3:4 and 4:3 directly, and stills can be generated in four variations at once, a single concept can be spun into a full placement set without redoing the creative work each time.
3. Reviving archival and low-resolution photographs: Old scans, screenshots and images exported at the wrong size are a persistent nuisance. The upscaler reconstructs edge and texture detail rather than smearing it, which makes previously unusable material publishable — with the honest caveat that badly damaged sources have limits.
4. Concept exploration for design and pitch work: Because different engines have genuinely different aesthetic instincts, running the same brief through several models is a fast way to explore a visual direction before committing. Industry comparisons note that Nano Banana Pro, from Google DeepMind, treats photorealism as its baseline while Seedream, from ByteDance, is stronger on stylised visuals and multi-image consistency — a split that is directly useful when you are deciding which look to pitch.
5. Short-form video for creators without an editing suite: For a creator who needs a five- or ten-second clip to open a video or fill a transition, generating motion from a still is dramatically faster than shooting and cutting. The output is standard MP4, so it drops into any editor or uploads directly.
6. Presentation and documentation visuals: Slide decks, help articles and internal wikis need illustrative images that are on-brand and not stock-photo generic. Generating them to spec — right aspect ratio, right colour palette, no licensing ambiguity — removes a recurring small friction from knowledge work.

How to use Image FX

1. Open the site and sign in with a Google account. Signing in is what unlocks the credit balance and the higher-quality engines; new accounts are granted a small allowance of free credits to try the paid models before committing.
2. Choose your workspace from the Tools menu — Text to Image, Image to Image, Text to Video or Image to Video — according to whether you are creating from scratch or transforming something you already have.
3. Write the prompt. The official guidance is to lead with a style declaration such as "in the style of watercolour" or "3D render", then specify lighting with terms like golden hour or dramatic shadows, then add composition notes such as centred or rule of thirds.
4. Select the model and the output settings: aspect ratio, resolution, batch size and, for video, clip duration. The credit cost updates with your model choice, so you can see what a render will cost before you commit to it.
5. Generate, then iterate. A standard text-to-image render costs 4 credits, returns in roughly 4 to 15 seconds depending on the model, and can produce four variations at once. Video takes longer — usually between 30 seconds and two minutes.
6. Refine and export. Send a promising still through the upscaler or the style-transfer workspace for a second pass, then download. Video downloads as MP4 at your chosen resolution.

Tips & Best Practices

• Draft on cheap models, finish on expensive ones. With a fourfold to fivefold spread between the least and most expensive image engines, the efficient pattern is to nail composition and prompt wording at 4 credits per attempt, then spend 20 credits once on the version you actually want.
• Treat the strength slider as a precision instrument. For image-to-image work the published guidance maps 20–40% to subtle adjustment, 40–60% to a moderate style shift and 60–80% to dramatic reimagining. Starting at maximum strength is the most common way to lose the composition you were trying to keep.
• Match the engine to the intent, not to its reputation. A photoreal product shot and a stylised illustrated banner reward different models. Since switching costs you nothing but a dropdown, run the same prompt through two engines before deciding which direction to develop.
• Upscale last, never first. Enhancement multiplies whatever is already in the frame, including artefacts and soft edges. Settle composition and style at normal resolution, then enlarge the final selection.
• Budget video deliberately. A single clip can consume more credits than twenty still images, so lock your concept with stills before spending on motion — storyboard cheaply, animate once.
• Write motion prompts as camera directions. Describing what the camera does — a slow zoom, an orbit, a pan — produces more controlled results than describing what the subject feels or means.
• Keep an external archive of anything you care about. Because history retention depends on your plan tier, downloading finished work as you go is the safe habit rather than trusting the platform's gallery as your only copy.

Who is Image FX for?

• Small e-commerce operators and marketplace sellers who need a steady stream of product visuals and short clips but have neither a studio budget nor a retoucher on call.
• Social media managers and content creators producing high volumes of platform-native assets across several aspect ratios each week.
• Freelance designers and agencies using generative drafts for early concept exploration, moodboards and client pitches before committing production hours.
• Small business owners without design staff who need presentable menus, banners, listings and announcements produced quickly and in-house.
• Marketing generalists at small teams who own the whole funnel and cannot wait on a creative queue for every asset.
• Educators and presenters assembling slide and course visuals that need to be specific rather than generically stock.
• Curious experimenters comparing model behaviour, for whom the ability to run one prompt across several engines behind a single balance is the main attraction.

Platforms

• Web application, browser-first: Everything runs in the browser: the Tools menu lists Text to Image, Image to Image, Text to Video, and Image to Video as web workspaces. There is no desktop installation and no local rendering — generation happens server-side.
• Mobile access through the browser: The interface is reachable from a phone browser, though heavy prompt iteration and file management are more comfortable on a larger screen. No dedicated iOS or Android application is listed in the site's own navigation or footer.
• Google account sign-in: Access to credits, higher-quality engines and saved history is tied to signing in with Google.
• File handling: Uploads accept JPG, PNG and WebP at up to 10MB per file. Images download in standard web formats; video exports as MP4 at 480p, 720p or 1080p, which plays anywhere without conversion.
• No public developer API is documented on the site, so the product should be treated as an interactive workspace rather than a component you can wire into an automated pipeline. (Based on public information)

Pricing & Plans

Access begins free. New accounts receive a small starting allowance of credits, and signed-in users also get one free 2x upscale per day, which is enough to evaluate output quality before paying anything. The free tier carries two meaningful constraints, though, and both are stated plainly rather than buried. Images produced on the free tier carry a watermark, and the free model does not save generation history, so paid access is what unlocks clean exports and an organized archive.

Paid tiers are published as Basic at $11.99 for 300 credits, Pro at $27.99 for 1,500 credits, and Max at $55.99 for 4,000 credits. Credits are the universal currency: a still image costs 4 to 20 depending on engine, an upscale costs 10 to 40 depending on factor, and a video clip costs 50 to 105 depending on model. One-time credit packages are also sold for people whose usage is occasional rather than monthly. Two commercial details deserve attention before subscribing. Unused credits do not roll over to the next cycle, and your balance resets to the plan allowance when a new billing period begins. And the company states plainly that it does not offer refunds, citing partner platform costs and the compute consumed by each generation — its published policy treats a sale as final once credits land in the account, with narrow exceptions for duplicate charges, credits that were never delivered and platform-side failures.

Alternatives

• Google's ImageFX (labs.google): The similarly named free tool from Google Labs, tied to Google's own Imagen models with an interactive prompt-chip interface. Free and polished, but single-vendor and without the editing, upscaling or video pipeline.
• Midjourney: The reference point for aesthetic quality in stylised image generation, with a mature community and deep style control. Stronger art direction, narrower scope — no built-in video or utility editing.
• Canva Magic Studio: Generative features embedded in a full design suite with templates, brand kits and layout. Better when the image is one element of a designed document; weaker on raw model choice.
• Adobe Firefly: Positioned around commercially safe training data and tight Creative Cloud integration, which matters most to enterprises with strict rights requirements and existing Adobe workflows.
• Runway: A video-first generative platform with genuine editing timelines and finer motion control. The better choice when video is the deliverable rather than an occasional extra.
• Direct model APIs: Going straight to a model provider gives the lowest per-unit cost and full automation control, at the price of building and maintaining the interface yourself.

Limitations & Considerations

• Credits expire with the billing cycle. Unused balance does not carry forward, so buying a large plan for occasional bursts wastes money. Match the tier to steady monthly consumption rather than to peak months.
• Refunds are effectively unavailable. The company states plainly that it does not offer refunds, citing partner platform costs and the compute consumed by each generation. Use the free allowance as your evaluation period, because after purchase there is no practical unwind.
• Video is expensive relative to images. The platform's own documentation puts the compute cost of video generation at roughly 31.5 to 32.5 times that of an image. Even on the Max plan the realistic monthly output is dozens of clips, not hundreds.
• Upload and processing limits are real. Source files are capped at 10MB in JPG, PNG or WebP. The upscaler currently handles a single image at a time with no batch processing, and heavily compressed or very blurry sources will still come out soft.
• Free output is watermarked and unsaved. For evaluation this is fine; for anything client-facing it is not, which makes the free tier a trial rather than a usable long-term option.
• Content restrictions are enforced. The terms prohibit NSFW generation outright and forbid attempts to bypass safety filters, moderation or rate limits. Accounts that try are subject to termination.
• Model availability is inherited, not owned. Because the value proposition is aggregating third-party engines, the roster and its pricing can shift when upstream providers change their terms. Treat any specific model's availability as current rather than permanent. (To be verified)
• Third-party review coverage is thin. At the time of writing there is no substantial independent review or community discussion tied to this specific domain, partly because several unrelated products share the name. Judge it on your own test renders rather than on reputation signals.

FAQ

Q1. Is Image FX free to use?

There is a genuine free tier. New accounts receive a starting allowance of credits, and signed-in users get one free 2x upscale each day. Two limits define it: images produced on the free tier carry a watermark, and the free model does not save generation history, so paid access is what unlocks clean exports and an organized archive. Treat free access as a proper evaluation window rather than a permanent plan.

Q2. Can I use the images and videos commercially?

Yes, on a paid plan. You retain full ownership of the images and videos you generate, and commercial use is included in the Basic, Pro, and Max plans while the free tier is limited to personal use. Paid output also comes without a watermark, which is usually the practical blocker for client work. As with any generative tool, you remain responsible for not generating material that infringes an existing trademark or likeness.

Q3. How does the credit system work?

Every action draws from one balance, priced by the compute it consumes. A standard text-to-image render costs 4 credits, returns in roughly 4 to 15 seconds depending on the model, and can produce four variations at once. Premium image engines cost more per render — up to 20 credits — upscaling costs 10 to 40 by factor, and video runs from 50 to 105 credits per clip. The important caveat is renewal behaviour: unused credits do not roll over to the next cycle, and your balance resets to the plan allowance when a new billing period begins.

Q4. Which AI models can I actually use?

On the image side the roster includes the platform's own ImageFX Pro engine plus Imagen 4, Seedream, Qwen Image, GPT Image 2 and Nano Banana Pro. Instead of paying for several separate services, you pick a model per job inside the same workspace, because each model excels at different styles. For motion, image-to-video runs on four models — Veo 3.1, SeeDance Pro, SeeDance Lite, and Wan 2.7 — priced between 50 and 105 credits per clip.

Q5. Is this the same as Google's ImageFX?

No. It is not the Google Labs tool of a similar name that lives on the labs.google domain, and the two products share nothing beyond a naming collision. This service is operated commercially under the Image-Fx AI and Image-Fx Labs name and runs a credit-based model. The overlap that causes confusion is that Imagen 4, a Google model, appears here as one selectable engine among many — availability of a model is not common ownership.

Q6. What are the file and output limits?

Uploads accept JPG, PNG and WebP files up to 10MB each. Images generate from 1K to 4K, and upscaled output is capped at 20,000 pixels on the longest side. Video exports as MP4 in 480p, 720p or 1080p, at five or ten seconds, across five aspect ratios. One workflow constraint worth planning around: the upscaler currently handles a single image at a time with no batch processing, and heavily compressed or very blurry sources will still come out soft.

Q7. Is my uploaded and generated content used to train AI models?

No, according to the published policies. The published policies state that the service does not use your generated content to train AI models — a commitment repeated in both the terms of service and the privacy policy. The privacy policy separately discloses that data is shared with payment processors, cloud hosting providers and analytics services, and that users in GDPR jurisdictions can request erasure and data portability.

Q8. Can I get a refund if the plan does not suit me?

Practically, no. The company states plainly that it does not offer refunds, citing partner platform costs and the compute consumed by each generation. Its refund policy treats sales as final once credits are issued, with narrow exceptions for duplicate charges, credits never delivered, platform-side technical failures and failed generations where credits were not automatically returned. Cancelling a subscription stops future billing but does not refund credits already issued for the current period.

Q9. Do I need to install anything, and is there a mobile app?

No installation. Everything runs in the browser: the Tools menu lists Text to Image, Image to Image, Text to Video, and Image to Video as web workspaces. A phone browser will load the interface, but no dedicated iOS or Android application appears in the site's own navigation. Sign-in is through a Google account, and generation happens server-side, so an older machine is not a constraint.

Q10. Are there content or account restrictions I should know about?

Yes, and they are explicit. Account holders must be at least 18 years old or the age of legal consent in their jurisdiction. The terms also prohibit generating NSFW material and forbid any attempt to bypass content safety filters, moderation checks or rate limits. Violations can lead to account termination, so these are worth reading before building a workflow on the platform.