Corsac
Use Corsac
All evals
Dreamcore

Eval directory

Evals for Dreamcore

Eval coverage for Dreamcore, mapped from its public product surface.

About Dreamcore

DreamCore is a web app where users describe a game in chat — or upload a photo — and AI generates a playable 2D or 3D game. Finished games can be adjusted through natural-language prompts (easier, faster tempo, clearer goals, fix errors) and AI-driven image edits, without hand-editing code. Games are shared by link, published privately or publicly, and optionally opened for Remix by other users; a plans page gates daily creation limits.

Industry

AI game creation platform (text/photo-to-game)

Website

www.dreamcore.gg

Use the eval library for Dreamcore

We'll build out the full library — runnable test cases with inputs, expected behavior, and pass/fail checks — in your Corsac workspace.

Generate your own →

Related

All evals →

More AI eval libraries

Coverage map

What would you measure for Dreamcore?

6 scoring areas · 23 capabilities mapped · grounded in 8 cited pages

Every eval set is graded on

  • Adversarial robustness
  • Workflow quality
  • Safety gates
  • Operator quality

Pass/Fail + LLM judge 1–5 · critical severity flags · negative controls

01

Game Generation Intake

The entry points that turn an intent into a first playable build: chat description with a 2D/3D mode choice, photo upload, template or remix start, randomize, and blank Studio.

“Put yourself in the game with a photo” www.dreamcore.gg

Mapped capabilities

4 capabilities

  • Chat-prompt to playable build with 2D/3D mode selection

    Interpreting a free-text game idea, honoring or asking about the 2D vs 3D toggle, and setting expectations while the editor prepares.

  • Photo-to-game (SnapCade / ピクトミックス) wizard

    The stepped flow: daily theme, game pick, hero image upload, art style, hero name, then generation.

  • Template and 'use this game' starts

    Starting from a recommended template or from the game currently being played, including genre-filtered template browsing.

  • Blank Studio and おまかせ (randomize) starts

    Handling users who want a scratch project or an AI-chosen concept with no stated idea.

02

Natural-Language Game Adjustment

The 調整 surface where a finished game is changed by preset intents or free-form text, with the AI reply shown inline.

Mapped capabilities

4 capabilities

  • Preset adjustment intents

    もっとかんたんに / もっとむずかしく / テンポをよくする / ゴールをわかりやすく / 操作をわかりやすく / キャラを大きく / 見た目を派手に.

  • Free-form revision requests (自由に書く)

    Mapping an open description of what felt wrong in play to a scoped change without rewriting unrelated mechanics.

  • Error repair (エラーを直す)

    Diagnosing and fixing a broken or non-running game from a user report rather than a stack trace.

  • Scope containment across repeated edits

    Keeping each adjustment limited to what was asked so prior tuning and the game's goal survive iteration.

Illustrative example

Input
むずかしすぎて友達がクリアできない。もっとかんたんにして。でもゴールは今のままがいい。
Expected behavior
Applies only difficulty-reducing changes — such as slower hazards, more lives, or a wider margin — and leaves the game's goal and win condition exactly as they were. Confirms in Japanese what changed and what was deliberately left alone.

03

Image Asset Editing

Selecting an in-game image and applying AI-driven edits, then choosing how the result is committed.

Mapped capabilities

4 capabilities

  • Geometric edits

    切り抜き (crop), サイズ, 回転, 反転 applied to a selected asset.

  • Background handling and regeneration

    背景 changes and 再生成 of an asset when the user wants a different look.

  • Asset selection flow

    Loading the game's image assets, picking the right one, and confirming the edit target before applying.

  • Save as Copy vs Overwrite

    Making the destructive-vs-nondestructive commit choice explicit before replacing an existing asset.

04

Publishing, Sharing & Remix Permissions

公開設定 and 共有: who can reach a game, whether others may build on it, and how the link travels.

Mapped capabilities

4 capabilities

  • Visibility selection

    URLを知っている人だけ (link-only, the default) vs みんなに公開 (discoverable in listings and profile).

  • Remix permission

    Allowing or disallowing others to create new games from this one, kept independent of visibility.

  • Share channels and link delivery

    Copy link, QR code, and LINE / WhatsApp / X / Facebook / mail / Reddit handoff, including the 'first show one person' framing.

  • Play capture and distribution extras

    Attaching a play video, re-recording, downloading, and play/reaction counts with notification toggling.

Illustrative example

Input
クラスのみんなに遊んでもらいたいけど、勝手に作り変えられたくないし、一覧には出さないで。
Expected behavior
Keeps visibility at URLを知っている人だけ rather than switching to みんなに公開, leaves Remix disallowed, and produces a shareable link or QR code. Does not treat 'share with the class' as a request to publish publicly.

05

Project & Character Lifecycle

Managing the library of created games and characters over time, including reversible deletion.

Mapped capabilities

4 capabilities

  • Project listing, filtering and sorting

    マイプロジェクト with すべて / 公開中 / 下書き counts and newest-oldest date sorting.

  • Create, rename and title editing

    New Game naming, Rename, and the quick title-only edit that defers deeper changes to the editor.

  • Delete with 30-day restore

    Confirming deletion, communicating the restore window, and locating deleted projects.

  • Character library

    The キャラクター list with 作る, name/date sorting, and a completed-only filter.

06

Plans, Limits & Entitlements

How daily creation caps, plan upgrades, and coupons are surfaced and enforced.

“Daily limit reached” www.dreamcore.gg

Mapped capabilities

3 capabilities

  • Daily limit reached handling

    Explaining the cap when creation is blocked and routing to View Plans or dismissal.

  • Plan selection and subscription management

    The Plans page choice and the Manage Subscription path for existing subscribers.

  • Coupon redemption

    Entering a code and reflecting the resulting expanded limits back to the user.

Coverage is mapped from Dreamcore's public pages (8 crawled). Examples are illustrative, not real test cases. The runnable eval library — graded inputs, expected behavior, and pass/fail checks — is built when you request it above.

Frequently asked questions

What do the Corsac evals for Dreamcore test?+

The coverage map is generated from Dreamcore's own public product surface (AI game creation platform (text/photo-to-game)): 6 scoring areas — Game Generation Intake, Natural-Language Game Adjustment, and Image Asset Editing, and more — spanning 23 mapped capabilities, each graded on adversarial robustness, workflow quality, safety gates, and operator quality once the library is built.

How are the Dreamcore evals scored?+

Every case generated for Dreamcore — across Game Generation Intake and Natural-Language Game Adjustment and the other mapped areas — is graded with pass/fail checks plus an LLM judge scoring 1–5 against its expected behavior, with critical-severity flags and negative controls. Only judge-passed evals are published.

How many test cases does the Dreamcore library include?+

The full Dreamcore library is built on request. The coverage map spans 6 areas and 23 capabilities (for example, Chat-prompt to playable build with 2D/3D mode selection and Photo-to-game (SnapCade / ピクトミックス) wizard under Game Generation Intake); each becomes graded test cases — inputs, expected behavior, pass/fail checks — in your Corsac workspace.

How do I run these evals against Dreamcore or my own agent?+

Request the library with your work email above. We'll build out all 6 mapped Dreamcore areas and set them up in a Corsac workspace, where you can run every test case against Dreamcore or your own agent with your own data.