PoseFlow

📷 Image placeholder

TODO: A hero composition showing PoseFlow in action, a phone running the consumer app with the green skeleton overlay tracking a user mid-squat on the left, and the Studio’s phases / measurements panel authoring the same movement on the right. Tells the “author once, run everywhere” story instantly.

Path: /images/hero/poseflow-overview.png

PoseFlow ships in two halves:

• The SDK (package:pose_flow), a Flutter plugin that wraps a pure-C pose engine and turns raw camera frames into rep counts, phase transitions, and form feedback.
• The Studio (package:pose_flow_studio), an embeddable authoring surface that lets a trainer build a new movement in front of a camera without writing any code. The output is a single .pose file that every consumer of the SDK can run.

What it does

┌─────────────────────┐    ┌─────────────────────┐    ┌──────────────────────┐
│  Camera frames      │ -> │  Pose detection     │ -> │  Pose landmarks    │
└─────────────────────┘    │  (PoseFlow's native runtime / WASM) │    └──────────────────────┘
                           └─────────────────────┘              │
                                                                v
┌─────────────────────┐    ┌─────────────────────┐    ┌──────────────────────┐
│  Studio (.pose)     │ -> │  MovementTracker      │ <- │  Per-frame state     │
│  - phases           │    │  - phase machine    │    └──────────────────────┘
│  - tracking points  │    │  - rep counter      │              │
│  - form rules       │    │  - form service     │              v
│  - rep detection    │    │  - bucket detector  │    ┌──────────────────────┐
└─────────────────────┘    └─────────────────────┘    │  Reps + form + cues  │
                                                      └──────────────────────┘

Out of the box you get:

• Pose tracking: 33-landmark pose output (front-facing camera, iOS / Android / web).
• Rep counting: anti-cheat validated counter that observes phase sequences and refuses partial-range reps.
• Form analysis: declarative FormRules evaluate per-frame and emit feedback events (cue text + severity + trigger).
• Camera-angle awareness: a 16-bucket detector classifies the user’s orientation (front / 45L / 90L / 45R / 90R × hip / floor / overhead + behind) so view-dependent measurements pick the right reference frame.
• A no-code authoring tool: the Studio gives a trainer a live camera, draggable phase tiles, and a measurement library. Saved movements are .pose files that any consumer of the SDK can run unchanged.

Read next

If you’re integrating the SDK:

• Getting started, install, scaffold, run your first movement in 30 minutes.
• SDK overview, architecture, hot paths, where to hook in.
• API reference, every public class.

If you’re authoring movements with the Studio:

• Studio overview, what it is, how it ships.
• Authoring workflow, phases → measurements → conditions → save → ship.
• Camera buckets, front / side / overhead variants.

If you’re integrating with both:

• Building a Studio host, wire the StudioPersistence ports to your storage backend.
• .pose file format spec, the file the Studio produces, byte by byte.

Status

• SDK: production. Rep counting, form analysis, and camera-angle detection are battle-tested across thousands of authored movements.
• Studio: production. Designed to embed in any Flutter host alongside the runtime SDK.