Unity
JuiceBox - AI Powered Animation
$50
Rated 5.0/5 from 14 ratings on Unity.
Price history
Recorded standalone sale prices and the latest available store price, in USD. Historical dots use the sale start date and its recorded offer price. Prices can change during a sale; this is not a daily price log. Bundle prices and upcoming offers are excluded. History covers up to the 100 most recent recorded offers.
Lowest recorded standalone price in this history: $25. The latest store price is $50.
View 5 recorded prices
| Sale start / price check | Price | Observation |
|---|---|---|
| $50 | Latest store price | |
| $25 | Unity New Release Discount 2026-08-26 2026-09-09 | |
| $25 | Unity New Release Discount 2026-08-26 2026-09-09 | |
| $25 | Unity New Release Discount 2026-08-26 2026-09-09 | |
| $50 | Unity First Price 2026-08-26 2026-08-26 |
Sale and bundle history
The 4 most recent recorded offers including this asset. Past offers may no longer be available. Bundle prices cover a package, not this asset alone.
About JuiceBox - AI Powered Animation
- Tools, Systems & Learning > Development Tools & Automation
Give AI coding assistants the documentation, tools, and animation language they need to effortlessly create and edit JuiceBox animations in Unity.
AI animation often feels off because the model cannot judge feel. It guesses numbers, and every correction is a full round‑trip. Code‑first tween libraries don’t help; the judgment lives outside the agent.
JuiceBox LLM improves the guess first. Instead of raw lerps, it offers named concepts—Swing, Follow, easing—so the first attempt lands much closer to the desired feel. The agent still can’t see the result, but it no longer has to invent feel from nothing.
Then it moves the remaining judgment to the human. Your agent builds a complete JuiceBox animation asset. You open it in the graph editor, drag the overshoot, and modify values live while the animation runs. The agent does what it’s good at; the person who can see the screen does what needs eyes.
This is AI animation you can edit, covering UI, cameras, GameObjects, effects, and gameplay feedback. It is not character animation or mocap generation.
Setup is simple:
1. Import the package. A setup window opens on first launch.
2. Click “Set up agent access” to configure Unity’s CLI or MCP server.
3. Describe the animation you want.
The package publishes JuiceBox’s entire animation vocabulary as typed, schema‑validated agent tools. It is not a chatbot bolted into the editor; generative AI works directly on your project without exposing a new API.
No API key, subscription, credits, or generation limits. JuiceBox LLM never calls an external model, so there are no billing concerns or keys to manage.
It works with the AI tools you already use. The integration is model‑agnostic and compatible with Claude Code, Cursor, GitHub Copilot, Claude Desktop, Unity’s AI Assistant, ChatGPT, Gemini, or any local LLM served through Ollama.
Natural language in, working animation out. Describe what you need in any AI chat, and the integration builds a real sequence that can be refined conversationally. The prompt engineering is already done: the package ships LLM‑readable contracts for every effect, parameter, easing, and behavior, an AI‑ready JSON animation format (.jbx) with a patch format (.jbxpatch) for targeted edits, and a machine‑readable manifest of the full vocabulary.
There is no code generation. Text‑to‑animation produces validated animation data that opens in JuiceBox’s visual editor, so nothing generated can break the build.
Safe by design. Every JuiceBox operation is a typed tool with a JSON schema, so an agent’s edit either matches the schema or fails loudly. Edits are validated and applied all‑or‑nothing, change tokens prevent overwriting hand edits, every change is undoable, and the file watcher only runs if you enable it. The editor is standalone; no LLM dependencies are added to builds.
Example requests:
- Animate this menu: slide the panels in, stagger the buttons, and add hover feedback.
- Add camera shake, hit flashes, and knockback reactions to this combat scene.
- Make pickups float, spin, highlight on proximity, and fly to the player on collect.
- Vibe code a subtle looping ambient motion with layered effects for my character.
- Make everything in this scene feel snappier and more impactful.
- Add natural variation so groups of objects don’t animate in lockstep.
The results are normal JuiceBox animations. Inspect, refine, and reuse them in the visual editor, or keep iterating by conversation. Nothing about them depends on an AI at runtime.
AI animation often feels off because the model cannot judge feel. It guesses numbers, and every correction is a full round‑trip. Code‑first tween libraries don’t help; the judgment lives outside the agent.
JuiceBox LLM improves the guess first. Instead of raw lerps, it offers named concepts—Swing, Follow, easing—so the first attempt lands much closer to the desired feel. The agent still can’t see the result, but it no longer has to invent feel from nothing.
Then it moves the remaining judgment to the human. Your agent builds a complete JuiceBox animation asset. You open it in the graph editor, drag the overshoot, and modify values live while the animation runs. The agent does what it’s good at; the person who can see the screen does what needs eyes.
This is AI animation you can edit, covering UI, cameras, GameObjects, effects, and gameplay feedback. It is not character animation or mocap generation.
Setup is simple:
1. Import the package. A setup window opens on first launch.
2. Click “Set up agent access” to configure Unity’s CLI or MCP server.
3. Describe the animation you want.
The package publishes JuiceBox’s entire animation vocabulary as typed, schema‑validated agent tools. It is not a chatbot bolted into the editor; generative AI works directly on your project without exposing a new API.
No API key, subscription, credits, or generation limits. JuiceBox LLM never calls an external model, so there are no billing concerns or keys to manage.
It works with the AI tools you already use. The integration is model‑agnostic and compatible with Claude Code, Cursor, GitHub Copilot, Claude Desktop, Unity’s AI Assistant, ChatGPT, Gemini, or any local LLM served through Ollama.
Natural language in, working animation out. Describe what you need in any AI chat, and the integration builds a real sequence that can be refined conversationally. The prompt engineering is already done: the package ships LLM‑readable contracts for every effect, parameter, easing, and behavior, an AI‑ready JSON animation format (.jbx) with a patch format (.jbxpatch) for targeted edits, and a machine‑readable manifest of the full vocabulary.
There is no code generation. Text‑to‑animation produces validated animation data that opens in JuiceBox’s visual editor, so nothing generated can break the build.
Safe by design. Every JuiceBox operation is a typed tool with a JSON schema, so an agent’s edit either matches the schema or fails loudly. Edits are validated and applied all‑or‑nothing, change tokens prevent overwriting hand edits, every change is undoable, and the file watcher only runs if you enable it. The editor is standalone; no LLM dependencies are added to builds.
Example requests:
- Animate this menu: slide the panels in, stagger the buttons, and add hover feedback.
- Add camera shake, hit flashes, and knockback reactions to this combat scene.
- Make pickups float, spin, highlight on proximity, and fly to the player on collect.
- Vibe code a subtle looping ambient motion with layered effects for my character.
- Make everything in this scene feel snappier and more impactful.
- Add natural variation so groups of objects don’t animate in lockstep.
The results are normal JuiceBox animations. Inspect, refine, and reuse them in the visual editor, or keep iterating by conversation. Nothing about them depends on an AI at runtime.