Krea AI Best real-time canvas
Real-time AI creative studio with a live canvas that renders photorealistic images in under 50ms as you draw, plus a six-model enhancer stack reaching 22K resolution and aggregated access to 40+ image and video generation models.
Where Krea came from
Krea was founded in March 2022 by Victor Perez and Diego Rodriguez, two Spanish engineers who wanted to solve a specific frustration: AI image generation felt disconnected from the creative act itself. Every tool at the time followed the same loop — write a prompt, press generate, wait three to ten seconds, decide if the result was close enough to keep, iterate. It was like having a conversation with a two-second delay on every sentence. Creative momentum died in the gap.
Their answer was the Realtime Canvas — a generation engine where the model runs continuously as you draw, sketch, or adjust sliders. The image doesn't render when you press a button. It renders while your mouse is moving. That distinction, which sounds small in description, changes everything about how the tool feels in use.
By 2025 Krea had raised approximately $83 million in funding and reached a reported $500 million valuation. The company remained small and product-focused — their model count, feature depth, and release cadence outpace what the team size would suggest. In June 2025 they launched Krea 1, their first proprietary image generation model, ending their reliance on third-party models for their flagship quality tier.
The platform has always skewed toward professional creative workflows rather than consumer novelty generation. Their user base maps heavily to concept artists, motion designers, architects, and brand designers — people for whom speed of iteration is the primary constraint, not the cost of API calls.
What Krea actually is
Krea is best described as a creative studio built around a real-time generation engine. That positioning distinguishes it from two other categories of AI image tool: the prompt-and-wait generators (Midjourney, DALL-E 3, Ideogram), and the workflow-automation platforms (ComfyUI, Automatic1111). Krea isn't primarily about prompting to a result, and it isn't primarily about building pipelines. It's about staying in direct visual contact with the AI while you work.
Five things ship under the Krea umbrella:
- Realtime Canvas — draw, upload, or sketch; the AI generates a continuous photorealistic interpretation at sub-50ms latency
- Image Generation — standard text-to-image with 20+ models including Krea 1, native 4K, 1000+ styles
- Enhancer / Upscaler — six image and six video enhancement models, upscaling to 22K resolution
- Video Generation — aggregated access to Veo 3, Kling, Hailuo, Runway, Sora, and more, plus video enhancement to 8K at 120fps
- LoRA Training — custom model fine-tuning on faces, products, or artistic styles, usable across the platform
The unifying logic is that all of these tools share one credit system, one interface, and one asset manager. You don't need five subscriptions to access the best model for each task. You pay Krea and pick the model that fits the moment.
The Realtime Canvas — Krea's standout feature
The Realtime Canvas is genuinely unlike anything else in the category. The interface splits the screen: left side is your input canvas — draw shapes with basic brush tools, drop in a reference photo, add colors, place text. The right side is the live render. Not a preview. Not a low-resolution thumbnail. A full-resolution photorealistic interpretation of whatever is on the left, updating continuously at under 50 milliseconds per frame.
The practical effect is that ideation becomes tactile. You sketch a rough silhouette of a chair. The right side shows a photorealistic chair matching your rough proportions. You make the legs longer with a single brush stroke. The rendered chair's legs update in real time. You add a color swatch. The chair recolors. You add a style prompt — "Bauhaus, studio lighting" — and the aesthetic shifts while preserving your composition. At no point did you press a button or wait for a generation.
For concept artists, this changes the early exploration phase fundamentally. The traditional workflow is: sketch in one tool, export, generate in another, compare, adjust, regenerate. The Krea workflow is: sketch, see, adjust, done. You can explore twenty compositional variants in the time a standard generator would complete three generations.
The sub-50ms latency comes from a distilled model architecture that reduces generation to 4–8 steps, then runs through compiled model graphs and hardware-optimized attention mechanisms. It's not running a full Flux inference on every brush stroke — it's a purpose-built model that sacrifices some absolute quality ceiling for continuous responsiveness.
The Realtime Canvas also supports uploading reference images as composition guides. Drop in an architectural rendering and sketch over it — Krea preserves your spatial structure while rendering in whatever visual style you specify. This is the specific workflow architects and interior designers have latched onto: 3D grey-box renders from SketchUp or Rhino become photorealistic client-ready visuals without touching Photoshop or waiting for a traditional rendering pipeline.
One honest note: the Realtime Canvas's output quality is lower than what Krea 1 or Flux generate in standard mode. The speed compromise is real. For final deliverables, you'll use the standard generator. For exploration, the canvas is the right tool and nothing else comes close.
The Krea 1 model
Krea launched its first proprietary image generation model in June 2025. Before Krea 1, the platform was entirely a front-end aggregator — excellent UX on top of third-party models. Krea 1 makes them a model company as well.
The positioning is deliberate: Krea 1 is built for artistic photorealism, not photographic accuracy. The distinction matters. The model is optimized to produce images that feel intentional — with crisp textures, controlled composition, and aesthetic range — rather than images that are technically correct at the expense of visual interest. Product photography, editorial looks, concept art, and architectural visualization are the stated sweet spots.
Generation speed is competitive: up to four images in approximately six seconds. Native resolution is 1.5K, upscalable to 4K through the built-in enhancer chain. The model natively supports LoRA fine-tuning, meaning custom trained styles apply to it directly. It also supports a Style Reference System — upload up to three images and the model uses them as aesthetic anchors, not as literal content to remix.
The model costs six compute credits per generation, mid-tier on Krea's credit scale. Compared to Flux Pro or Midjourney v6 for the same creative use case — stylized, compositionally strong photorealism — Krea 1 holds its own. It won't win a strict photographic realism comparison against a fine-tuned Flux checkpoint, but for the iterative creative workflow Krea is built around, the quality ceiling is appropriate.
One limitation to flag: Krea 1 is explicitly not the model to use for precise anatomical outputs or highly literal technical renders. It interprets prompts with an artistic eye. If you need exact hand geometry or specific technical accuracy, route to a different model in the picker.
Enhancer and Upscaler — the depth chart
The Krea Enhancer is one of the most capable upscaling stacks available without buying a standalone tool. Six image enhancement models ship under one subscription, each with a different character. Understanding which to reach for is the skill that separates power users from people who just hit 4x on everything.
The six image enhancer models
Krea Enhance is the go-to for upscaling AI-generated outputs. It's tuned specifically for the artifacts and softness common to generated images, and it generates new detail rather than just interpolating. Output ceiling is 8K; generation time around 30 seconds. This is the first choice when you generate in Krea and want to push to print quality.
Upscale V1 is the fastest option (about 5 seconds) and the most neutral. It preserves what's already there rather than adding AI-generated texture. If you have a clean original and just need it physically larger, V1 is correct. If your original is soft or has AI smearing artifacts, it will preserve those too.
Bloom is the creative upscaler. It's faithful to the original but adds texture density and clarity in a way that reads as naturally photographic. The 60-second generation time is the tradeoff. Good for mid-resolution source images you want to land at high-quality editorial-print resolution.
Topaz and Topaz Generative are the heavy options — Topaz reaching 22K, Topaz Generative reaching 16K with more aggressive AI texture synthesis. These are the choices when you need wall-size print output or when the source material is badly degraded and you want the AI to make confident reconstruction choices. Advanced controls (strength, resemblance, texture density, denoise, face detection) give professional-level tuning depth.
Krea Legacy is the original model, still useful for its Scene Transfer capability — which can apply a visual style to the upscaled output rather than just enlarging what's there. Ceiling is 4K, which makes it more limited than the others, but the style transfer hook has no direct equivalent in the other models.
For video, a parallel set of models (Proteus, Iris, Artemis, Nyx, Gaia, Apollo) handles video enhancement up to 8K and frame interpolation up to 120fps. Krea is one of very few platforms where video upscaling of this caliber is included in the same subscription as image generation.
For the highest quality large-format output, chain two upscale passes — 2x then 4x — rather than going straight to 8x. The incremental approach gives each model more structured input to work with and produces fewer hallucinated textures at the edges of detail areas.
Video generation — aggregated access
Krea's video generation is not a single model. It's a unified interface over the leading third-party video models: Google's Veo 3, Kling (multiple versions), Hailuo, Runway, Wan, and Sora 2. You pick the model, write the prompt or upload a reference frame, and generate. The outputs are managed inside Krea's asset system alongside your images.
The practical advantage is context switching speed. If you're doing a product shoot in Krea — image generation, enhancement, delivery — you can move directly to a short product video without opening a second browser tab. The Krea credit system handles billing across all video models. Per-second pricing is roughly $0.10 for Sora 2 and $0.07 for Kling 2.6 (paid out of your plan credits), which is standard market rate for those models.
Krea adds its own video layer on top: frame interpolation to 120fps, video upscaling to 8K via the enhancement pipeline, lipsync support, and video restyling. These capabilities bridge the gap between "generate raw video" and "deliver finished video" without leaving the platform.
Video generation is credit-heavy. On Pro at $35/mo (20,000 credits), you get roughly 37 Kling videos at standard settings. If video is your primary use case, run the math before committing: Krea's Pro tier delivers less video volume per dollar than a standalone Kling subscription at equivalent pricing. Krea's advantage is the workflow — not the per-video cost.
LoRA training — custom style models
Krea's training pipeline lets you fine-tune a custom LoRA on your own images — faces, products, architectural styles, character designs. Once trained, the LoRA applies across Krea's generation toolset: standard generation, Krea 1, and video workflows can all pull from it.
The workflow is upload-and-caption: provide 15–30 images, write captions or let Krea auto-caption, set the trigger word, run training. Training time varies with dataset size but typically runs in the 20–40 minute range for a basic style model. The output is a LoRA file stored in your Krea account, accessible from any generation prompt via the model picker.
For product and brand work this is the feature that makes Krea a professional tool rather than a consumer one. A LoRA trained on a specific product makes every subsequent generation — images, videos, 3D renders — consistent to that product's visual identity without prompt engineering heroics each time.
Max plan is required for LoRA training that goes beyond the basic tier (2,000 training images vs the Basic/Pro cap). For teams doing ongoing brand work, this is the primary reason to evaluate Max over Pro.
Multi-model aggregator angle
Krea's model list in 2026 covers more ground than any direct competitor. The image side includes Krea 1, Flux Kontext and its variants, Ideogram, stable diffusion checkpoints, and twenty other models selectable per generation. The video side covers the full market: Veo 3, Kling, Sora 2, Hailuo, Runway, Wan. The enhancer adds Topaz models that would otherwise require a separate Topaz Gigapixel license.
The aggregator angle solves a real problem: the best model for a given creative brief changes. Krea 1 for editorial photorealism. Flux Kontext for image editing and regional modification. A fine-tuned stable diffusion model for a specific illustrative style. Topaz for print upscaling. Without Krea you'd pay for each separately and context-switch between interfaces constantly.
The trade-off is depth. Midjourney gives you one model and an extremely well-tuned prompt experience for that model. Using Flux via Replicate directly costs less per generation. Topaz Gigapixel as a standalone app has more upscaling controls than Krea's Topaz integration. Krea trades best-in-class depth on any single capability for best-in-class breadth across all of them in one interface.
For a creative professional who uses image generation, enhancement, and video across a week's work — which is most of Krea's target audience — paying one subscription and opening one tab is worth the trade.
Three real workflows, end-to-end
New product, no photography budget, client needs 12 hero images
Starting point: three clean product photos on a white background. Goal: twelve hero images across different contexts — kitchen, office, travel — without a studio booking.
Step one: train a LoRA on the product using the three source photos. Krea auto-captions and builds the model in 25 minutes. The LoRA attaches to the product's specific geometry and material finish, so subsequent generations maintain brand fidelity without prompt gymnastics.
Step two: use the Realtime Canvas to sketch compositions for each context. Rough perspective lines on the left; the right side shows the product in a photorealistic kitchen setting. Adjust proportions live — slide the product left, the lighting reads differently in real time. Lock a composition that feels right in about two minutes per scene.
Step three: generate each scene at full quality via Krea 1 with the LoRA active. Six credits per image. Step four: Bloom enhancer to 4K at print quality. Total credits used: modest against a Pro plan. The twelve images took under three hours, including LoRA training time. A studio day for equivalent coverage would have been a four-figure booking.
Grey-box 3D model needs to become a photorealistic client deck
The scenario: a small residential project, SketchUp grey-box model, client presentation in two days. The brief is eight render variants showing different material palettes — brick, white render, wood cladding — across two lighting conditions (day, dusk).
Export eight reference frames from SketchUp as flat renders. Drop each into Krea's Realtime Canvas as reference inputs. Add style prompts per material — "red brick, mortar joints, warm afternoon light." The canvas renders a photorealistic facade interpretation matching the composition of the grey-box while applying the material vocabulary.
Each variant takes about four minutes to land on the right visual — sketching variations on the canvas, adjusting the style prompt, locking the composition. Then a Topaz 4x upscale to presentation resolution. Eight variants, a morning's work. No V-Ray license required, no overnight render queue.
The honest limitation: Krea handles the mid-fidelity "is this the right direction?" presentation well. For a planning submission or construction document, you still need traditional rendering. Krea is the right tool for the exploratory phases and for client approvals, not for the final technical drawings.
Low-resolution archival scan needs to become printable
Starting point: a 600px x 400px scanned archival photograph, medium-quality JPEG, slight compression artifacts, soft edges throughout. Requirement: printable at A2 size, approximately 5000px x 7000px output.
First pass: Krea Enhance 4x. The model reconstructs facial detail and edge sharpness specific to AI-generated patterns rather than photographic noise. Result at 2400px x 1600px is noticeably sharper but still under target.
Second pass: Topaz 2x with face detection enabled, resemblance set to 0.85, denoise at moderate, texture generation moderate. Topaz adds grain-level texture density while the face detection pass keeps portrait detail controlled. Output at 4800px x 3200px is print-ready for A2 at standard DPI.
Total credits used: mid-range. Time: about eight minutes of active work, thirty minutes of processing. A comparable Topaz Gigapixel workflow would produce similar results but requires a separate $199/year license and a desktop install. Krea's result is slightly lower ceiling on absolute control, but the integration convenience is significant for users who aren't running dedicated image processing machines.
Performance, measured
Across representative generation tasks comparing Krea 1 against Midjourney v6 and Flux Pro for stylized photorealism, and the Krea Enhancer stack against Magnific for upscaling quality, here's how the platform positions:
bench --tool=all --metric=realtime,quality,upscale,value image generation tools 2026
The summary: Krea wins on breadth and on the unique real-time capability. It loses on raw cost-per-generation when compared to either a focused competitor or a direct API. The right way to read the comparison is not "which is cheapest per image" but "which saves the most context-switching time."
Who Krea is for — and who should look elsewhere
Reach for Krea if you are
- a concept artist who spends the first hour of any project in rapid visual exploration — the canvas eliminates the generate-wait-adjust loop entirely
- a product or brand designer who needs consistent visual identity across image, video, and 3D without managing five separate tools
- an architect or interior designer who converts CAD and 3D renders into client-facing photorealistic visuals regularly
- a small agency creative who needs the full model stack — image generation, video, upscaling, LoRA training — on one invoice
- anyone who currently pays separately for Topaz Gigapixel and an image generator — the Krea Pro math likely saves money
Look elsewhere if you need
- maximum volume generation at the lowest cost per image — Flux via Replicate API or Leonardo on their token plans will produce more images per dollar
- the absolute highest quality ceiling for a single model — Midjourney's depth of prompt tuning for its own model is still unmatched for that specific workflow
- enterprise support with SLAs, email ticketing, and account managers — Krea's support is Discord-first, which doesn't suit every procurement environment
- purely text-based or document workflows — this is a visual tool front to back
Where Krea falls short
Support is Discord-only
This is the most consistent complaint in Krea's user feedback across 2025 and 2026. There is no email support, no in-app ticket system, and no live chat beyond the Discord community. For independent creators this is workable. For teams who've had billing issues — and billing disputes with Krea are documented in multiple reviews — a Discord server is an insufficient resolution path. Krea needs a proper support tier before it can credibly serve business clients at scale.
Credit-to-output value is below competitors
Direct comparison: at the Pro tier ($35/month, 20,000 credits), Krea produces roughly 37 Kling-quality videos or approximately 3,300 standard image generations (6 credits each via Krea 1). Leonardo at a comparable price point produces substantially more images per dollar. If you're doing high-volume generation rather than thoughtful iteration, Krea's credit economy penalizes you. The platform is built for quality-per-session, not volume-per-dollar.
Enhancer quality has degraded post-update
Multiple users have noted that an enhancement model update in late 2025 reduced the quality of certain upscaling results — specifically Krea Enhance on photographic portraits. The original Krea Legacy model is still available as a fallback, and Topaz models are unaffected, but it's a genuine rough edge in a toolset that markets heavily on enhancement quality.
Realtime Canvas output ceiling
The canvas generates at display resolution in real time. The quality is excellent for exploration but not deliverable-grade without a follow-on standard generation pass. New users sometimes mistake the canvas output for a final result and are surprised by the quality difference when they export. The workflow is: explore on canvas, lock composition, generate final via Krea 1 or Flux, then enhance. That's two to three steps, not one.
Free tier is more restricted than it appears
The 100 daily compute unit limit sounds reasonable until you realize that a single Krea 1 generation costs 6 units and a Topaz 4x upscale costs substantially more. A meaningful creative session on the free tier will hit the ceiling within an hour. The free tier is adequate for evaluating whether you like the platform, not for doing real work. Plan for Basic ($9) at minimum if you intend to use it regularly.
Pricing, in real terms
Krea's credit system makes pricing less transparent than a flat monthly cap, but the structure is straightforward once you know the per-task costs.
Free ($0/mo) gives 100 compute units daily, no credit card required. Enough to try every feature and evaluate quality. Not enough for a real project session.
Basic ($9/mo) gives 5,000 monthly credits, commercial license, full image access, and 4K upscaling. For a designer doing light to moderate image work — under 800 generations per month — Basic is the right tier. Entering the platform at $9 is genuinely competitive for what's included.
Pro ($35/mo) gives 20,000 credits, all video models, Nodes automation workflow tool, and 8K upscaling. Video generation becomes available here. For creative professionals using both image and video pipelines, Pro is the functional minimum. Worth the jump from Basic if you use video even occasionally.
Max ($70/mo) gives 60,000 credits, 22K upscaling, unlimited concurrency, and up to 2,000 LoRA training images. The right tier for production studios doing ongoing brand work or for anyone doing heavy enhancement of large-format content. The concurrency feature — running multiple generations in parallel without queue wait — is the hidden value at this tier for teams with time pressure.
Business ($200/mo) adds 80,000 credits, team seats up to 50, credit rollover, and analytics API. The team structure is notable: $200 covers the whole team, not $200 per seat. For a five-person creative team, that's $40 per person — cheaper than three Pro accounts.
Annual billing shaves approximately 20% across all plans. The lifetime deal has appeared occasionally on platforms like AppSumo but isn't a permanent fixture — worth watching if budget is the primary constraint.
Krea vs Magnific AI
a/krea b/magnific-ai
Magnific AI is the specialist upscaler — it built its reputation on aggressive AI hallucination of missing detail, particularly in skin, fabric, and architectural texture. Krea and Magnific overlap heavily on the enhancer side, but serve it from different platform contexts.
krea wins at
- full creative suite — generation, video, training in one sub
- 22K ceiling vs Magnific's ~16K
- video upscaling to 8K — Magnific is images only
- lower entry price ($9 vs Magnific's $39 start)
- real-time canvas with no Magnific equivalent
magnific wins at
- upscaling depth and fine-grain control
- community-documented prompt recipes for specific looks
- dedicated support vs Discord-only
- more predictable credit-to-upscale pricing
Verdict: If upscaling is your primary task and volume matters, Magnific's dedicated toolset is slightly sharper. If you already use Krea for generation, the built-in enhancer stack eliminates the need for a separate Magnific account for most use cases.
Krea vs Leonardo AI
a/krea b/leonardo-ai
Leonardo AI is Krea's closest direct competitor as a multi-model platform — also freemium, also supporting custom LoRA training, also aggregating multiple generation models. The differences are in depth versus breadth and in workflow philosophy.
krea wins at
- real-time canvas — Leonardo has no equivalent
- 22K upscaling vs Leonardo's 4K ceiling
- video model breadth (Veo 3, Sora 2, Kling)
- 3D generation included in platform
- team billing — one plan, up to 50 members
leonardo wins at
- value per credit — more images per dollar
- image quality consistency across models
- established community and template library
- better documented model-specific workflows
Verdict: Leonardo if your priority is high-volume generation at better per-image cost. Krea if real-time iteration and advanced upscaling are part of your daily workflow. Both are credible; they serve slightly different creative rhythms.
Krea vs Flux
a/krea b/flux
Flux by Black Forest Labs is the most capable open-weight image generation model family available in 2026 — Flux Pro, Flux Kontext, Flux Dev. You can access Flux via Replicate, via fal.ai, or via Krea itself. The comparison is more nuanced than it looks.
krea wins at
- real-time canvas (runs its own distilled model)
- integrated enhancement and video in one UI
- no-code LoRA training vs Flux's technical setup
- asset management across all outputs
- accessible pricing for non-technical users
flux wins at
- raw generation quality ceiling
- cheapest cost per image via direct API
- open-weight — no platform dependency
- fine-tuning depth and model control
Verdict: Krea actually includes Flux models in its picker — so framing it as either/or misses the point. Use Krea's interface for the workflow convenience and tap Flux when you need its specific quality character. Going direct to Flux via API only makes sense if you're comfortable with technical setup and want maximum control at minimum cost.
FAQ
Is the Realtime Canvas output good enough to deliver to clients?
For concept presentations and direction approvals, yes. For final production assets, no — the canvas runs a distilled model for speed that trades some quality ceiling. The correct workflow is: explore in the canvas, then generate the locked composition via Krea 1 or Flux at full quality, then upscale via the enhancer chain. Treat the canvas as an ideation tool, not a delivery tool.
Does Krea own my images?
No. Any plan with a commercial license (Basic and above) gives you full ownership and commercial rights to your generated outputs. The free tier watermarks outputs and restricts commercial use. Read the current terms carefully if you're in a regulated industry — standard AI image ownership questions apply to Krea outputs as they do everywhere.
How does LoRA training work and what data does Krea use?
You upload your training images to Krea's servers for the training process. After training, the LoRA is stored in your account. Krea's terms state they do not use your uploaded training data to train platform models. For sensitive brand or personal images, this is worth reading carefully before proceeding — the images do leave your local machine.
Can I use Krea offline?
No. Krea is a web platform. All generation runs on Krea's servers, not locally. An iPad app with Apple Pencil support exists for mobile use, but it requires an internet connection. There is no offline mode.
Is Krea 1 the best model on the platform?
It's Krea's best proprietary model and the best for artistic photorealism. But "best" depends on the task. Flux Kontext beats Krea 1 for regional editing and image modification. A well-tuned Flux checkpoint may outperform Krea 1 for a specific aesthetic. The platform is designed for you to swap models per task — Krea 1 is the default starting point, not the only answer.
What happened to Krea's enhancer quality after the late 2025 update?
Multiple users reported degraded portrait enhancement quality from the Krea Enhance model post-update. The Topaz and Topaz Generative models were unaffected. Krea Legacy (the original model) is still accessible if you need the pre-update behavior. Krea acknowledged the issue in their Discord. As of early 2026, a fix was in progress but not yet shipped for all affected workflows.
How does Krea compare to Midjourney for pure image quality?
Midjourney remains the reference point for prompt-to-stunning-image in a single shot. Its aesthetic tuning and community-developed prompting depth are hard to match. Krea 1 is competitive for professional photorealism but serves a different workflow — iteration-heavy, composition-first. If you want the fastest route from one strong text prompt to one remarkable image, Midjourney is still the choice. If you want to explore and refine visually, Krea wins.
Is the Business plan really one fixed price for 50 people?
Yes, as of the current pricing structure — $200/mo covers up to 50 team members. At five people that's $40 per person, cheaper than three Pro subscriptions. At ten people it's $20 per person, roughly half the per-user cost of Pro. For teams of three or fewer, individual Pro accounts are probably simpler. For four or more, Business math works in your favor.
The verdict
The most tactile AI image tool available. Not the most economical.
Krea's Realtime Canvas is the most genuinely novel interaction model in AI image tooling right now. Nothing else lets you draw and watch a photorealistic image build itself in under 50 milliseconds. If you do any work where visual exploration speed is the constraint — concept art, product design, architectural visualization — Krea changes the shape of that work in a way that a faster prompt-and-wait tool cannot match.
The platform is broader than the canvas alone: Krea 1 for final generation, a six-model enhancer stack that reaches 22K resolution, aggregated access to every major video model, and no-code LoRA training. One subscription replaces three to five single-purpose tools for most creative professionals.
The deductions against a perfect score are real: credit value is below competitors for high-volume generation, Discord-only support won't scale for business teams, and the enhancer quality regression from late 2025 hasn't fully recovered. The platform rewards iterative creative work over volume work. Understand what you're paying for and it's excellent value. Try to use it as a cheap bulk generator and you'll leave disappointed.