Stock vs Custom Illustration: A Scene-Coverage Audit

Executive Summary

Stock vs custom illustration is a scene-coverage decision. Start with what each image must explain, then check whether an existing asset communicates that message in a compatible visual language. Choose permitted adaptation when a defined change closes the gap. Commission original work when the necessary content remains missing, or use an accurate product screenshot, a diagram or no illustration when that communicates more directly. GOV.UK's design guidance supports illustrations that simplify complex subjects; it does not make decoration a requirement. ^[1]

Free access, customization and originality are different properties. unDraw permits commercial use under its license, while Storyset permits use and modification subject to its terms. Neither permission establishes that a candidate depicts a particular product workflow. ^{[2] [3]} Adobe explicitly states that its Stock licenses are nonexclusive. Conversely, commissioned artwork does not automatically transfer copyright or provide unrestricted exclusivity: Illustrators Australia describes exclusivity in relation to the uses specified in the written agreement. ^{[4] [5]}

The quantitative evidence should inform a quote request, rather than become a universal price chart. Retrieved 31 October 2026, Adobe's US-locale page displayed US$29.99 per month for an annual plan billed monthly with 10 Stock credits per month, including up to 10 images, vectors, music or templates. ^[6] Upwork displayed an illustrator range of US$15 to US$30 per hour, explicitly based on certain historical contracts worldwide. That is a platform reference, not a current quote, a representative market survey or a basis for inventing production hours. ^[7]

This report's fictional RelayDesk launch demonstrates an auditable sourcing worksheet using exact Storyset candidate pages in the Bro style. The pages' labels and download options were checked; the artwork itself could not be visually inspected in this research environment. Accordingly, candidates remain unapproved, and no scene-coverage percentage or library ranking is reported. Buyers should record that distinction rather than count a search hit as a usable image. Accessibility also affects the choice: color cannot be the sole carrier of meaning, and Carbon recommends text alone for an empty state when space is limited. ^{[8] [9]} The practical outcome is a mixed sourcing plan with explicit gaps, a narrow adaptation request and a custom brief only for work that is actually needed.

Introduction and Background

A SaaS launch may need a homepage image, an explanation of a new feature and a support or success state. These are different communication tasks even when they share a palette. A generic collaboration scene might introduce a team-oriented service. It cannot be accepted as evidence of how that service assigns work merely because its title contains a relevant keyword. W3C's image guidance distinguishes images that convey concepts or information from images whose purpose is decoration. That distinction provides a more useful starting point than choosing an illustration supplier first. ^[10]

The familiar comparison already has substantial practitioner coverage. Studio Giraffe recommends beginning with purpose and message. Anchor & Alpine describes a semi-custom route between stock and original illustration. (Source: www.studiogiraffe.sg) ^[11] The additional method here is to make that decision inspectable: name the required scene, preserve an exact candidate URL, identify the mismatch and ask what work would remove it.

Scene coverage means coverage of the messages in a particular brief. It does not mean catalog size, visual quality in general or the popularity of a library. Visual fit is an editorial judgment about whether a candidate suits those messages and the buyer's visual constraints. A different audience or brief can produce a different judgment without either assessment being dishonest.

The intended readers are marketing leads, founders and agency creative buyers planning websites, sales decks and recurring campaign series. The publisher, Tapflare, occupies the custom-production side of this discussion; its appearance in the comparison is disclosure of that position, not independent evidence for the method. No publisher pricing is used.

The central question is therefore specific: can this source explain the required things with an acceptable amount of permitted work? Atlassian's guidance that illustration should reinforce rather than replace the message is a useful boundary. An image that needs extensive verbal correction may be the wrong candidate, however attractive it looks in isolation. (Source: atlassian.design)

Stock Illustration

Capabilities

Stock illustration is existing artwork selected for a new application. In this report, it includes free libraries as well as licensed commercial catalogs. Buyers can inspect available subjects before requesting original production, but should distinguish the library's browsing features from the suitability of any individual scene.

ManyPixels exposes named styles including Playstroke, Birdview, Chromablue, Colossalflat and Azureline, together with category browsing and color controls. These are useful search constraints, not proof that a particular style covers a launch brief. ^[12] unDraw describes color generation for matching a brand, and its exact Teamwork page offers Scalable Vector Graphics (SVG) and Portable Network Graphics (PNG) downloads. ^{[13] [14]}

A repeatable search starts with the message and expands through related actions:

• Task: Write what the viewer should understand after seeing the image. (Source: www.studiogiraffe.sg)
• Action: Search for the activity, such as assigning, reviewing or resolving.
• Objects: Record essential items that must appear, rather than accepting a matching title.
• People: Specify necessary roles and representation without treating characters as interchangeable decoration. ^[15]
• Style: Keep a named family or a documented visual treatment constant during the first pass. ^[1]
• Evidence: Save the item URL and the reason for accepting or rejecting it.

Adoption

Use stock when an existing scene can perform the required job without implying unsupported product behavior. A supporting collaboration image and a product explanation should have separate rows in the audit, even if they appear on the same page.

The sourcing check also includes the actual license. unDraw permits commercial use but excludes uses such as assembling a similar or competing service or distributing illustration packs. This is a narrower permission than unrestricted redistribution. ^[2]

Strengths and Limitations

The practical strength is a visible candidate that can be evaluated before an original-art brief is commissioned. The limitation is whatever remains different between that candidate and the required message. Neither is a universal statement about production speed or business performance.

Do not stretch a generic scene into a brand identity asset without checking the applicable permission. Adobe, for example, states that Stock images cannot be used in or as a logo. That is Adobe's documented limitation, not a claim that every library has identical terms. ^[4]

Keep no illustration available as an intentional choice. For a compact support or empty state, a clear sentence and action may carry the message sufficiently. Carbon explicitly recommends text alone where space is limited. Buying an image is unnecessary when the layout has no useful job for it. ^[9]

Adapted Stock Illustration

Capabilities

Adapted stock starts with an existing asset and changes it within the applicable permission. ManyPixels includes modification in its gallery license, while Storyset's terms also grant a qualified permission to use and modify its content. These permissions should be read alongside restrictions, rather than treated as a transfer of the original work. ^{[12] [3]}

Documented customization controls can help identify feasible changes. Storyset describes hiding elements through its layer list and selecting colors through a picker or RGB and Hex values. These controls establish particular operations; they do not demonstrate that an arbitrary scene can be rebuilt into an accurate product explanation. ^[16]

The adaptation request should name the operation:

• Palette: Identify the actual colors to change and the colors that carry meaning.
• Composition: Specify the crop, spacing or placement needed for the intended layout.
• Elements: List any objects to remove, move or add, subject to file inspection.
• Typography: State whether labels belong in the illustration or in surrounding page text.
• Variants: Identify required contexts, such as a narrow feature card and a wide slide.
• Boundary: Say which changes would create a new scene rather than adapt the existing one.

Adoption

Choose adaptation when the underlying concept is already suitable and the remaining work is bounded. “Match the teal palette and remove an irrelevant background object” is a more quotable request than “make this feel like the brand.” The latter leaves both the scale of the change and the acceptance criterion unclear.

Recoloring alone does not establish consistency. GOV.UK's guidance includes line widths and border radiuses as well as color. IBM's illustration system also describes consistent shapes, angles and radii. These are examples of original design-system rules, not universal prescriptions for every brand. ^{[1] [15]}

Strengths and Limitations

Adaptation can preserve a useful starting composition while addressing a specific mismatch. Its limit is reached when changes alter the central action, relationships or story. At that point, obtain a separate original-work quote rather than assuming the library file makes the whole task an edit.

Adobe permits editing non-editorial Stock assets subject to its stated restrictions. That permission should not be generalized to editorial assets or to another supplier's license. ^[4] Buyers should also ask whether the available source actually exposes the objects that need changing.

Color changes must preserve understandable meaning. W3C's guidance recommends additional cues, such as shape or text, when color conveys information. A green success mark that becomes brand purple still needs an understandable label or symbol if its meaning matters. This is a communication requirement for the adaptation, not merely a palette preference. ^[8]

Custom Illustration

Capabilities

Custom illustration begins with the required message and a commissioned composition. Its strongest sourcing justification is an unresolved need: a particular relationship between people and objects, a product-specific process, or a visual family that must support future scenes.

The selection of an illustrator still matters. Illustrators Australia recommends approaching artists whose established style suits the project. A custom commission is not a promise that any illustrator can produce any treatment. ^[5] Atlassian similarly identifies unique stories around major launches or complex features as situations where custom illustration may be needed, within its own design guidance. (Source: atlassian.design)

For actual-product explanation, supply approved references and distinguish literal facts from optional metaphor. ^[17] An illustrator should not have to infer a workflow from a landing-page slogan. If an accurate screenshot communicates the feature more directly, the original work may instead be limited to annotations or a simplified explanatory diagram.

Adoption

A useful brief describes the missing content before listing deliverables. IllustrationX asks for composition, dimensions, quantity, style, layouts, schedule, usage and budget when quoting. This helps separate the drawing problem from assumptions about where the work will appear. ^[17]

A compact scene brief can contain:

• Message: The single idea the scene must communicate.
• Audience: The reader and the context in which the image will appear.
• Required content: Essential roles, objects and relationships.
• Product accuracy: Approved screens, names and behavior that must remain true.
• Visual treatment: Style references, line and shape rules, palette and representation.
• Candidate history: Exact stock URLs and the specific gaps they leave.
• Quote scope: Original work, adaptations, variants and the intended usage.
• Review milestone: What the buyer will approve at rough stage before final rendering.

References should communicate direction rather than request imitation. The Association of Illustrators advises against copying another illustrator's work or imitating their style. Describe the characteristics the project needs and choose an artist capable of developing a suitable approach. ^[18]

Strengths and Limitations

Custom work can address the brief's missing concepts directly. It also requires decisions that a stock title cannot resolve: what should happen in the scene, which relationships matter, and how literal the depiction should be.

Use a rough-stage review to check those decisions before final artwork. IllustrationX describes rough approval as the opportunity to provide further direction; Illustrators Australia recommends agreeing the number and timing of rough sketches and included changes. These are sourced process examples, not a universal revision allowance. ^{[17] [5]}

Finally, originality and ownership are separate questions. The Graphic Artists Guild's model agreement reserves rights not expressly granted to the artist. It is a model to adapt, not the agreement for every commission. Confirm the intended uses in the actual proposal rather than assuming payment means unrestricted ownership. ^[19]

Feature Comparison

A sourcing route should be evaluated against the same communication task. (Source: www.studiogiraffe.sg) Table 1 compares options by what must be established before buying production. It deliberately avoids a per-image price ranking, because the underlying scopes differ.

The table supports a mixed decision. A team can source a generic supporting image from a library, adapt a suitable companion scene and commission the product explanation. It need not choose a single route for every placement, although the resulting family still needs coherent direction.

Before comparing proposals, define the visual constraints:

• Style family: Name the source family or describe the intended treatment.
• Geometry: Record relevant shape, line, perspective and detail rules.
• Palette: Separate decorative colors from colors with semantic meaning.
• Representation: Specify roles and contexts appropriate to the audience.
• Working format: Identify what must remain editable for the proposed changes.

These constraints should be concrete enough to inspect. GOV.UK's consistency guidance provides a useful example of measurable visual properties. ^[1] File extensions are weaker evidence: MDN notes that SVG can contain bitmap images, so an SVG label alone does not prove that every object is separately editable. ^[21] IBM's guidance to consider gender, cultural and geographical diversity likewise shows why representation belongs in the brief rather than being left to whichever search result appears first. ^[15]

Performance and Benchmarks

Scene-Coverage Audit (Hypothetical Example)

RelayDesk is fictional. Its launch brief has three communication tasks: introduce shared teamwork, explain its actual assignment-rule workflow, and reassure users that support is available. These requirements are invented solely to demonstrate the sourcing method. They are not a case study, customer result or market benchmark.

The candidate sample is deliberately small. The exact Storyset pages opened were Remote team, Pricing plans and Business support, each in the Bro style. Their item pages expose SVG and PNG download options and license summaries. ^{[22] [23] [24]} These labels establish candidate identity and availability, not what every character or object looks like.

Research limitation: page text was accessible, but the artwork files could not be visually inspected after the built-in fetch and permitted fallback failed. The audit therefore records metadata checks and outstanding visual decisions. No candidate is certified as usable, and no absence is generalized to the whole catalog.

Table 2 records the candidate-level evidence and the remaining quote questions. A buyer repeating the audit should replace every unchecked cell after inspecting the artwork.

The product row records a lack of verified coverage, not an exhaustive failed catalog search. The other rows are candidates, not acceptable scenes. This prevents a misleading result in which finding three relevant titles becomes “complete coverage.”

For a live audit, use these states consistently:

• Usable as found: The inspected image communicates the task and satisfies the stated constraints.
• Adaptation needed: The underlying scene is suitable and the necessary permitted changes are listed.
• No suitable result: The checked candidates do not meet the requirement; record exactly what was searched.
• Not checked: Search, visual inspection, format inspection or permissions remain incomplete.

Do not force all states into a sample to make it look finished. The state should describe the evidence actually obtained.

Product Accuracy and Message Tests

A useful comparison asks what the viewer can learn from the image. (Source: atlassian.design) For the fictional assignment-rule scene, show the proposed image beside the feature statement. Ask whether the relationship between the rule and the outcome is understandable, or whether the viewer merely sees generic business activity. This is a suggested buyer review, not a validated comprehension benchmark.

A screenshot can be an appropriate marketing-page candidate when the interface itself is the subject. Context matters: Atlassian recommends low-fidelity interface depictions and discourages actual screenshots specifically inside its products, because of the screen-within-screen effect. That advice should not be generalized to every sales deck or website. (Source: atlassian.design)

If the explanation is complex, provide its essential information in text as well. W3C recommends making complex-image descriptions available to everyone. ^[25] For screenshot and line-art exports, MDN recommends lossless encoding to avoid conspicuous artifacts around sharp edges and text. Inspect the actual result at its intended display size rather than assuming a source-format label guarantees legibility. ^[21]

Data Analysis and Evidence

What the Available Prices Establish

The price evidence is a dated purchasing reference, not an illustration-market average. On 31 October 2026, Adobe's US-locale page displayed US$29.99 per month, on an annual plan billed monthly, for 10 Stock credits per month, including up to 10 images, vectors, music or templates. The asset allowance and annual commitment belong beside the price. It should not be represented as an unrestricted single-image purchase. ^[6]

Upwork displayed US$15 to US$30 per hour for illustrators, describing the underlying data as certain historical contracts worldwide. The fetched page did not establish a representative sample size or measurement period for a general market estimate. ^[7] Upwork also states that actual payment is negotiated with the talent. A buyer therefore needs a project quote and should not multiply that range by assumed hours to manufacture a custom-image price. ^[7]

IllustrationX identifies audience size among the factors influencing cost, alongside complexity, licensing, usage duration and territory in its commissioning guidance. That makes a shared usage brief necessary when comparing quotations. ^[17]

Table 3 separates documented inputs from amounts the buyer must obtain. It is a cost worksheet, not a claim that one route is always cheaper.

The comparison should use equivalent required scenes and applications. ^[17] A free download with an unresolved product explanation is not the same purchased outcome as a complete, accurate diagram. Equally, an original-art quote should not be assigned to a supporting image that already works.

A Buyer-Specific Calculation

Use the following accounting structure with actual quotes:

Total sourcing cost = applicable license or access cost + quoted adaptation work + quoted original work.

For a time-based adaptation quote, calculate its cost using the quoted hours and rate. If a supplier provides a fixed fee, use that fee instead. Record buyer search time separately rather than assigning an invented value to it. Allocate subscription spending explicitly if the subscription also serves unrelated work; otherwise, a scene-level comparison can hide its assumptions.

Keep a change boundary in the quote. The Association of Illustrators advises charging for significant alterations outside the original brief. A changed product story should therefore trigger a scope discussion rather than silently being counted as another minor revision. ^[18]

Coverage Is a Local Measure

After inspection, an optional coverage fraction is acceptable scenes divided by required scenes for that named brief and source sample. Define whether “acceptable” includes approved adaptations or only assets usable as found. Report critical uncovered concepts beside the fraction, since a missing product explanation can matter more than several available decorative scenes.

For RelayDesk, there are three required communication tasks and three candidate pages, but no completed visual approvals. No coverage fraction is calculated. This is an incomplete sourcing audit with disclosed evidence limits, not a measured result for Storyset or stock libraries generally. No fetched evidence establishes a general conversion, revenue or engagement advantage for any sourcing route.

Implications and Future Directions

The useful future asset is the buyer's own scene inventory: a record that can be extended when the next feature, audience or campaign arrives. Reuse the requirements and candidate history, then reconsider whether the existing family can express the new message. Do not assume that a successful supporting scene proves coverage of a later product story.

The inventory should also carry the intended accessibility treatment. W3C recommends a null text alternative for purely decorative images. An informative image has a different job and needs an alternative appropriate to its content and purpose. ^[10] For complex images, W3C describes a short identification plus a longer textual representation of the essential information. Accessibility therefore belongs in the sourcing decision, while the message and composition are still being chosen. ^[25]

Before issuing a request, complete this short sequence:

• Inventory: List the communication tasks and distinguish actual-product explanation from generic context.
• Inspect: Open exact candidate assets and record what remains unchecked.
• Constrain: Write the visual properties and editable elements that matter.
• Choose: Select stock, bounded adaptation, original work, a screenshot or no image for each task.
• Quote: Request equivalent usage, variants and a rough-stage review.
• Extend: Keep the scene record when adding later work, rather than restarting from catalog popularity.

Formats should follow the task. SVG can scale smoothly, according to MDN, but a team still needs to inspect its structure and the rendered result. ^[21] A compact empty state may need only text, consistent with Carbon's guidance, while a complex explanatory figure may need a more substantial textual companion. ^[9]

Preserve the source permission with each candidate without turning the audit into a legal handoff ledger. Storyset's terms contain restrictions on reselling, assigning, transferring or sublicensing content and derived work. ^[3] Adobe separately requires licensing the same asset for each client use. ^[4] For agency buyers, these are reasons to check the actual intended application before promising an adaptation. They are not reasons to assume that free assets are unusable or that original commissions automatically confer every right.

The result should be a focused production request: the scenes already covered, the defined modifications and the missing concepts to quote. Scheduling dependencies and approval roles can be addressed in the wider campaign plan without replacing this earlier sourcing decision.

Conclusion

Choose stock illustration when an inspected asset communicates the required message and fits the stated constraints. Choose adapted stock when the scene is suitable and a defined, permitted change closes the gap. Choose custom illustration when the missing content justifies original composition. An accurate screenshot, explanatory diagram or clear sentence can be the stronger choice when it addresses the communication task more directly. The purpose remains simplification and understanding, consistent with GOV.UK's illustration guidance. ^[1]

The scene-coverage audit makes this decision reviewable. It records what the image must communicate, whether actual-product detail is essential, which exact candidate was considered, why it falls short and what a supplier should quote. It also preserves honest unknowns. A relevant title is a candidate, a customization control is a capability, and an accepted scene is a separate conclusion.

The fictional RelayDesk example stops at that boundary: its candidate pages are traceable, but its visual decisions remain incomplete. It produces neither a library ranking nor a performance claim. A live buyer can finish the same worksheet with inspected artwork and project-specific quotations.

Commissioning does not remove the need for a clear agreement about intended use, just as a stock license does not establish suitability. Illustrators Australia's licensing guidance ties exclusivity to the written agreement's uses. ^[5] Apply the same discipline to the creative decision: define the message, inspect the evidence, request only the missing work and approve the route that explains the product accurately.