Skip to content
SiteList

Synth Review: capable product, weak technical health (53/100)

Synth scores 53/100 for a clear AI game-asset proposition, strong accessibility and usability signals, and a polished dark product surface.

Reviewed by SiteList Engine · 11 of 13 dimensions · published Reviewed on October 8, 2026

Is this your site?

Claim it

Quick facts

Synth is a usage-based AI game-asset product with free daily credits, free downloadable libraries, and paid plans that include commercial rights.

Field Evidence
Domain app.synth.run
Category AI game asset generation and free game-asset library
Pricing Usage-based; supplied price range: ?–? USD
Pages crawled 22
Crawl date 2026-10-08
Evidence
Pages crawled
22
Crawl date
2026-10-08

Executive summary

Synth communicates a concrete product quickly: it generates game assets from prompts for game developers and creators. Usability and accessibility are the strongest public signals, scoring 78/100 and 84/100; the supplied accessibility evidence also gives the site a Lighthouse score of 94. Design is coherent and polished, with no measured horizontal overflow at 390px.

The main weaknesses are structural. Technical SEO scores 18/100: all 22 crawled routes use a two-hop cross-domain redirect chain and canonicalize to one landing page, while a nonexistent-URL probe returns 200. Risk and stability scores 12/100 on related traffic-risk signals. Editorial QA scores 36/100 because titles and meta descriptions are duplicated across the crawled pages and route intent diverges from delivered content.

Evidence
Overall score
53/100
Technical SEO score
18/100
Accessibility score
84/100

01 · First impressions & positioning — clear game-asset job, split brand identity

Synth presents a clear product job: generate 2D assets, sprite animations, 3D models, and skyboxes from prompts, with exports for game engines. The identity is less clear. The audited host is app.synth.run, while the title, H1, logo link, canonical, and navigation identify Rosebud AI and point to rosebud.ai. The /company and /pricing samples also reuse the homepage metadata and canonical. Clarify the Synth/Rosebud relationship, choose one public identity, and give retained routes unique metadata.

Evidence
Crawled host
app.synth.run
Canonical
https://rosebud.ai/ai-game-assets

02 · Audience & messaging — useful capability map, but the first choice is diluted

The homepage communicates four generator categories and supports a broad audience, but it does not prioritize one first task. Visitors see 2D assets, sprite sheets and animation, 3D models, skyboxes, CREATE, PLAY, PRICING, SIGN IN, engine compatibility, credits, and a large FAQ. Cross-domain branding makes the next step harder to predict, while /company and /pricing repeat homepage content. Keep the category overview, then lead with one generation path and segment secondary routes by goal.

Evidence
Generator categories
4: 2D assets, sprite animation, 3D models, skyboxes
Route duplication
/company and /pricing repeat homepage metadata/content

03 · Usability — direct generator paths, with choice density and 15.8 s mobile interactivity

The supplied walkthrough scores usability 78/100: visitors can identify the offer, open a generator, and reach pricing in one navigation step. The homepage provides direct actions for the 2D and sprite-sheet generators, while the mobile view clearly states the game-asset proposition. Choice density is the main friction because generators, libraries, tutorials, games, and external destinations share the same surface. Mobile PSI reports 2.9 s LCP, 563 ms blocking time, and 15.8 s to interactive. Make generator actions dominant and defer secondary work.

Evidence
Usability score
78/100
Mobile time to interactive
15.8 s

04 · Accessibility — 94 Lighthouse score, but 82 /play inputs lack names

Synth has a strong accessibility baseline, with expected landmarks, English language declaration, no missing image-alt attributes in the sampled pages, and a Lighthouse accessibility score of 94. The critical exception is /play: the rendered DOM reports 82 inputs and all 82 lack a label or ARIA name. Lighthouse also flags a restrictive viewport that may prevent zoom. Add an accessible name to every user-facing input, remove restrictive zoom settings, and add a reduced-motion override for the 15 animation declarations found in CSS.

Evidence
Lighthouse accessibility
94
Unnamed /play inputs
82 of 82

05 · Design execution — coherent dark surface, but 64 mobile targets are undersized

Synth has a coherent dark, neon-accented game-asset surface, a clean heading outline, one H1, and no measured horizontal overflow at 390px. Mobile execution needs a focused pass. The capture reports 64 interactive elements below the 44px tap-target threshold; CREATE and PLAY measure 62×29px, while PRICING and SIGN IN measure 60×36px. Body copy is 13px, list items 12px, and buttons 12px. Remove maximum-scale=1, enlarge hit areas with padding, and raise mobile text sizes.

Evidence
Undersized interactive elements
64 at 390px
Mobile paragraph size
13px

06 · Performance — 2.19 s field LCP, dragged by 4.5 MB and 15.8 s lab interactivity

The homepage made 127 requests totaling 4,525 KiB; rosebud.ai contributed 3,117 KiB, with 682 KiB of unused JavaScript estimated. Split route code, defer non-essential analytics, and audit the cross-domain dependency before tightening budgets.

Evidence
Field p75 LCP
2.19 s
Homepage request set
127 requests / 4,525 KiB

07 · Writing quality — a concrete hero buried under duplicated 1,410-word pages

The strongest line is “Make AI game assets from a prompt, no design skills needed.” It names the job and audience directly. The crawl then repeats the same H1, metadata, headings, and 1,410-word body across 22 routes, including pricing, company, and a sprite-sheet article. Feature labels also render as “MakesCharacter sprites,” “You getPNG,” and “Works inUnity.” Give each route topic-specific copy and repair the spacing, paragraph boundaries, and image alternatives.

Evidence
Repeated page body
1,410-word paragraph on each of 22 pages
Broken labels
MakesCharacter / You getPNG / Works inUnity

08 · Decision-support surfaces — pricing is present, but plan fit is not defended

The pricing route is relevant, but the supplied evidence does not show a recommendation for a specific buyer situation or a complete cost matrix. A visitor needs to know which plan fits generation volume, when commercial rights matter, and what the cheaper option gives up. Show billing cadence, included credits, commercial-rights boundaries, and overage or renewal implications. Add recommendations for solo creators, small teams, and commercial projects, each with one named tradeoff and an overridable choice.

Evidence
Decision-support score
45/100
Pricing evidence
No complete plan/price matrix or defended recommendation supplied

09 · Risk & stability — 22 routes collapse through redirects and a soft 404

Risk is structural rather than a measured traffic-loss claim. All 22 crawled routes use the same two-hop chain and canonicalize to rosebud.ai/ai-game-assets, which can consolidate pages intended to rank independently. A guaranteed-nonexistent URL returns 200 at the same landing page. Implement one-hop, path-preserving redirects, use self-canonicals for retained pages, and return a genuine 404 for unknown paths. Then validate coverage in GSC over four to eight weeks.

Evidence
Redirect pattern
All 22 routes use the same two-hop chain
Nonexistent URL status
200

10 · Editorial QA of content — duplicated intent makes the route set not publish-ready

The current route set should not pass editorial QA. The crawler reports the same title and meta description on 22 pages, with repeated H1 and heading trees across /pricing and the sprite-sheet blog route. That article opens with “AI Game Asset Generator & Free Game Assets” instead of a sprite-sheet answer, so its lede fails. The crawl also exposes “You getPNG,” a 1,410-word paragraph per page, six images without alternatives, and two-hop redirects. Add a render-based gate for route intent, metadata, text boundaries, alt coverage, and final URLs.

Evidence
Duplicate metadata
Same title and meta description on 22 pages
Blog lede
Generic generator H1 on sprite-sheet article route

11 · Technical SEO — 18/100, with universal canonical collapse and a 200 soft 404

Technical SEO is the clearest weakness. Every crawled route follows app.synth.run → beta.pixelvibe.com → rosebud.ai/ai-game-assets and reports the same canonical target, collapsing distinct product, pricing, blog, and company intents into one landing-page signal. The nonexistent-URL probe also returns 200, confirming a soft 404. Publish a valid sitemap, replace the two-hop chain with path-preserving one-hop redirects, assign distinct self-canonicals, and return 404 for unknown paths. Raw and rendered extraction both measure 1,426 words, so rendering is not the primary blocker.

Evidence
Technical SEO score
18/100
Raw vs rendered extraction
1,426 words in both

Verdict — 53/100: useful product surface, serious technical risk

Synth is a plausible asset-generation surface for game developers and creators, but its public technical foundation needs repair before the product can communicate reliably through search and route-level discovery.

What works is clear: usability scores 78/100, accessibility scores 84/100, and the dark product surface scores 68/100 for design execution. The priority weaknesses are the 18/100 technical SEO result, the 12/100 risk and stability result, and the 36/100 editorial QA result. Together, they point to redirect and canonical consolidation, soft-404 risk, duplicated metadata, and route-content mismatch.

The product is best suited to creators evaluating AI-generated game assets and free downloadable libraries, with commercial-rights plans available for paid use. Fix the route and metadata structure first, then resolve the confirmed accessibility issue on /play.

Evidence
Overall score
53/100
Risk & stability score
12/100
Editorial QA score
36/100

Methodology & data notes

The byline describes this as a 13-dimension review.

Google Search Console is not connected, so no GSC enrichment is included. The supplied evidence gives no price range beyond ?–? USD. Read How SiteList scores for the scoring method and interpretation.

Evidence
Dimensions reviewed
13
GSC connected
false

Questions buyers actually ask

Who is Synth for?

Synth is aimed at game developers and creators who need sprites, animations, 3D models, skyboxes, and related assets for engines and tools including Unity, Godot, GameMaker, Phaser, Construct, Roblox Studio, Blender, or Unreal.

What does Synth offer?

The supplied profile describes AI game-asset generation and a free game-asset library, with free daily credits, free downloadable libraries, and paid plans that include commercial rights.

How accessible is Synth?

Synth scores 84/100 for accessibility, with a Lighthouse accessibility score of 94. The /play route has a confirmed issue: all 82 detected inputs lack an accessible name.

What should Synth fix first?

Prioritize the two-hop cross-domain redirect chain, the universal canonical, and the soft-404 risk. Then address duplicated titles and meta descriptions across the 22 crawled pages and the 82 unnamed /play inputs.

How this review was made

SiteList reviewed synth.run on October 8, 2026 — pages, screenshots, performance runs, structured data and public records — then scored it across 11 of the 13 public dimensions. Every claim above is sourced from what we collected; nothing is hand-tuned and the score is never for sale.

Not covered in this write-up: Review-content integrity (not applicable), Docs & self-serve help (not assessed). Dimensions without a score are excluded and their weight is redistributed across the scored ones.

Pending enrichment (data we could not fetch this run): Brand SERP samples could not be fetched because no SERP endpoint was configured., Live branded-search visibility and query ownership remain unverified., Task-based usability testing with game developers., Manual interaction testing of CREATE, PLAY, PRICING, SIGN IN, generator forms, and credit-cost disclosure., Live click execution and form validation were not available., Rendered pricing plan details were not present., Rendered company/customer-proof content was not present., live axe-core run and scripted tab-order probe…

Read the full methodology

53/100SynthJump to review