diff --git a/README.md b/README.md index 4f7670a..0411a09 100644 --- a/README.md +++ b/README.md @@ -3,15 +3,13 @@

Kinematic motion as text.
- Mermaid gave LLMs a way to draw diagrams. -
- Posecode gives them a way to show movement. + An inspectable, editable movement format for animation tools, LLMs, and web products.

- A human-readable spatial DSL for describing, validating, and rendering + Like Mermaid for movement: small text documents for describing, validating, and rendering
- exercises, physiotherapy movements, posture, dance, and human motion. + deterministic human motion without hiding the source inside a black box.

@@ -41,7 +39,9 @@ ## Why Posecode? -Ask an LLM to explain physical movement and it usually returns unstructured prose or a static diagram. +Animation clips and generated trajectories can show movement, but they often hide the semantic decisions that produced it. + +Posecode keeps those decisions in readable source. A human can write the document, an animation tool can emit it, or an LLM can draft it. Parsing, validation, editing, and rendering do not require an AI model.

@@ -75,14 +75,14 @@ A human may understand that instruction, but a renderer cannot reliably determin - in what sequence, - or within which physical limits. -Large language models can often reason about the components of human movement, but they lack a standardized syntax for expressing that reasoning in a renderable and testable form. +Humans, tools, and language models all need a shared syntax for expressing movement in a renderable and testable form. Posecode provides that missing representation. ### See Posecode in 28 seconds -From an LLM prompt to editable Posecode, validated 3D rendering, MCP tools, -and a one-script web embed. +From readable movement source to validated 3D rendering, MCP tools, and a +one-script web embed.

@@ -845,7 +845,7 @@ Feedback and contributions are welcome. ---

- LLMs already have languages for software, data, and interfaces. + Animation tools, LLMs, and web products need movement they can inspect.
- Posecode gives them a language for movement. + Posecode makes movement readable source.

diff --git a/playground/for-products.html b/playground/for-products.html index 84c231f..60f48e8 100644 --- a/playground/for-products.html +++ b/playground/for-products.html @@ -3,14 +3,14 @@ - Use Posecode in Your Product | Open-Source Motion Tools - + Integrate Inspectable Movement | Posecode for Products + - - + + @@ -18,8 +18,8 @@ - - + + @@ -50,17 +50,17 @@

For product owners & developers

-

Put movement inside your product.

-

Posecode turns a small, readable text document into an editable 3D movement. Use the open-source pieces now—or help shape the product infrastructure that should come next.

+

Build with inspectable movement.

+

Posecode is a readable movement format with a parser, typed IR, Three.js renderer, web component, and LLM tools. Author it directly or with an LLM, then keep the result editable and under your control.

Discuss an integration → See what exists today
@@ -68,7 +68,7 @@

Open-source toolkit today.
No hosted commercial platform yet.

Ways to integrate

Start at the layer you need.

-

Keep the standard player, compose the lower-level libraries, or give a local agent movement tools.

+

Keep the standard player, compose the lower-level libraries, or give an LLM movement tools.

@@ -92,7 +92,7 @@

Own the interface

- 03 / agents + 03 / LLMs

Run movement tools locally

The npm MCP server runs over stdio on your machine. It teaches an MCP client the language, validates documents, and creates playground links.

npx -y posecode-mcp@latest
@@ -109,11 +109,11 @@

Run movement tools locally

Movement that stays inspectable.

-

Animation editors

Generate a first pass from text, expose timing and joint values, and let a human revise the movement instead of accepting a black-box clip.

+

Animation editors

Load readable movement source, expose timing and joint values, and let a human revise the result instead of accepting a black-box clip.

Fitness & rehab products

Show an exercise beside a program or cue. Joint angles are clamped to general range-of-motion limits, but Posecode is not clinical advice or patient-specific validation.

Education tools

Connect written instruction to a moving figure for anatomy, dance, physical education, or movement vocabulary.

Movement libraries

Store small, diffable text files, validate them in CI, and open any movement in the browser without pre-rendering video.

-

Agent workflows

Let an MCP-capable local agent write, validate, repair, and return a link to movement the user can inspect.

+

LLM workflows

Let an MCP-capable LLM write, validate, repair, and return a link to movement the user can inspect.

Other web products

Use movement as editable content anywhere a video or static diagram is too rigid and a full animation pipeline is too heavy.

@@ -138,8 +138,8 @@

The current open-source surface.

Exploring next

-

Useful infrastructure should be shaped by real products.

-

We are interested in working with a small number of design partners to learn which operational layer actually matters. Possibilities—not announced products—include:

+

Real integrations should decide what comes next.

+

We are interested in working with a small number of design partners to learn which operational layer actually matters. Possible areas to explore include:

-

None of these capabilities is generally available today. There is no published commercial pricing. A design-partner conversation is for scoping the problem, constraints, and a possible collaboration—not selling a service that already exists.

+

None of these capabilities is generally available today. There is no published commercial pricing. A design-partner conversation is for scoping the problem, constraints, and a possible collaboration. It is not a sales call for a service that already exists.

diff --git a/playground/index.html b/playground/index.html index 2aac6a2..e21d3de 100644 --- a/playground/index.html +++ b/playground/index.html @@ -3,23 +3,23 @@ - Posecode: Describe Human Movement as Text, Animated in 3D + Posecode: Inspectable Human Movement as Text @@ -31,7 +31,7 @@ @@ -66,7 +66,7 @@ "@id": "https://www.posecode.org/#website", "url": "https://www.posecode.org/", "name": "Posecode", - "description": "An open-source text-to-motion language for LLMs. Describe human movement as text; render it as an animated 3D figure in the browser." + "description": "An open-source, inspectable movement format for animation tools, LLMs, and web products. Describe human movement as text, validate it, and render it in the browser." }, { "@type": "SoftwareApplication", @@ -75,11 +75,11 @@ "url": "https://www.posecode.org/", "applicationCategory": "DeveloperApplication", "operatingSystem": "Any (web browser)", - "description": "Posecode is a small language capable LLMs can use to describe single-person movement. The browser validates a closed vocabulary, constrains joint targets to configured ranges, and renders an inspectable animated 3D figure.", + "description": "Posecode is a small, readable format for single-person movement. Humans or LLMs can author it; the browser validates a closed vocabulary, constrains joint targets to configured ranges, and renders an inspectable animated 3D figure.", "offers": { "@type": "Offer", "price": "0", "priceCurrency": "USD" }, "license": "https://github.com/posecode-dev/posecode/blob/main/docs/legal/LICENSING.md", "sameAs": ["https://github.com/posecode-dev/posecode"], - "keywords": "text to motion, LLM, kinematic motion, 3D animation, exercise, physiotherapy, open source" + "keywords": "movement format, kinematic motion, editable animation, 3D animation, animation tools, LLMs, web products, open source" }, { "@type": "FAQPage", @@ -90,15 +90,15 @@ "name": "What is Posecode?", "acceptedAnswer": { "@type": "Answer", - "text": "Posecode is an open-source text-to-motion language. You describe single-person movement as phases, joint targets, timing, and contacts; the browser validates the document and renders it as an animated 3D figure." + "text": "Posecode is an open-source, inspectable movement format. A document describes phases, joint targets, timing, and contacts; tools can validate, edit, diff, and render the same movement." } }, { "@type": "Question", - "name": "How do I generate an exercise animation with an LLM?", + "name": "Does Posecode require an AI model?", "acceptedAnswer": { "@type": "Answer", - "text": "Copy the authoring guide into a capable chat model and ask for a representable movement such as a hip hinge or hamstring stretch. For a representable request, the model should return raw .posecode text. Paste it into the playground: valid documents animate, while parse errors, range clamps, and unresolved reach targets appear as diagnostics." + "text": "No. Posecode documents can be written by hand, produced by an animation tool, or drafted by an LLM. Parsing, validation, editing, and rendering are deterministic and do not require an AI model." } }, { @@ -138,99 +138,159 @@ For products GitHub - Try it live → + Try it live →
-
-

A motion language for LLMs

-

Give your LLM
a body.

-

- Give your LLM the Posecode prompt. Posecode turns its response into structured, editable 3D human - motion in your browser. -

-
- Open the playground → - +
+
+

A readable format for human movement

+

Movement you can inspect.

-
- Open source - Runs on-device - Prompt included +
+

+ Posecode turns human movement into structured, editable text for animation tools, LLMs, and web + products. Change the source, validate it, and render the same motion in the browser. +

+
+ Open source + motion → + Read the format +
+
+ Open source + Deterministic + Human or LLM authored +
-
-
- After pasting the Posecode prompt -

Then ask: “Write a controlled superhero landing.”

-
- +
- - -
01 -
LLM-native

Readable movement, not matrices.

+
Deterministic

Same document, same validated motion.

02 -
Range-aware

Targets stay within configured ranges.

+
Inspectable

Joints, timing, contacts, and warnings stay visible.

03 -
Browser-native

Live 3D, no server-side diffusion.

+
Built to integrate

Parser, renderer, embed, and LLM tools.

How it works

-

One prompt. Three clear steps.

+

From source text to usable motion.

  1. 01 -

    Ask

    -

    Describe a representable movement.

    +

    Describe

    +

    Write phases, joint actions, timing, travel, and contacts as readable text.

  2. 02 -

    Generate

    -

    A capable LLM writes a tiny .posecode file.

    +

    Validate

    +

    The parser produces typed movement data, clamps configured ranges, and reports diagnostics.

  3. 03 -

    Move

    -

    The browser validates, constrains, and animates it.

    +

    Use

    +

    Render it in Three.js, embed a player, or pass the same document through your own tools.

+
+
+

LLM authoring is optional

+

Write Posecode directly or start with an LLM draft.

+

+ A human can author the source, an animation tool can emit it, or an LLM can draft it. The result is + still a small document you can read, edit, diff, validate, and render without model inference. +

+
+
+
+ 01 / direct +

Write the movement

+

Use the playground, language tooling, or your own editor.

+
+
+ 02 / assisted +

Draft with an LLM

+

Give an LLM the authoring guide, then inspect and revise its output.

+ +
+
+
+

Movement library

-

Ask for almost any move.

-

Explore launch-ready examples across fitness, rehab, dance, and more.

+

Start from readable source.

+

Open examples across fitness, rehab, dance, and more. Every movement remains editable text.

@@ -238,42 +298,40 @@

Ask for almost any move.

For developers

-

Open all the way down.

-

Use the Apache-2.0 open standard or build with the AGPL-3.0 product layer.

+

Build on the layer you need.

+

Adopt the format, compose the runtime packages, or integrate the player into a product.

- - Build -

Playground →

-

Edit and preview movement live.

- -
- Connect -

MCP server

-

Let agents validate movement.

-
- Author + Define

Language spec →

-

A small, semantic grammar.

+

The portable movement contract.

+ + + Validate +

Parser + typed IR →

+

Turn text into constrained movement data.

+ + + Render +

Three.js runtime →

+

Preview procedural or compatible GLB figures.

 - - Extend -

Open-source packages →

-

Apache specification and core tools, AGPL product runtime.

+ + Integrate +

Embed + LLM tools →

+

Add inspectable movement to a web product or workflow.
-

Your turn

-

What should a body do?

-

Ask your LLM. Paste the result. Watch it move.

+

Make movement inspectable

+

Keep the movement as readable source.

+

Open the source beside the movement, change one instruction, and see exactly what changed.

- Start with the playground → - + Start with the playground → + Explore integrations →
@@ -288,7 +346,7 @@

What should a body do?