Get the betas

Two desktop apps, one pipeline. FullField Stitch turns two overlapping recordings into a single panoramic match video, and BallTrack AI turns a wide match video — ideally that panorama — into broadcast-style footage that follows the ball. Each works on its own; together they go from two tripods to TV-style highlights. Both are free while in beta — the trade is honest feedback: your benchmark numbers and your bug reports.

Step 1 · Stitch the match

FullField Stitch

Turns two overlapping 4K 60fps recordings of the same field — shot on GoPro or similar action cameras — into one seamless panoramic match video: audio-clap sync, GPU-accelerated stitching, hardware encoding, optional color LUTs. The panorama is a finished match video in its own right, and the ideal input for BallTrack AI below.

Coming soon

Windows beta installer

The Windows x64 installer is being packaged for the first beta wave. There’s nothing to download yet — this card will turn into a real download button when it ships.

Download for Windows — coming soon

Beta notes: closed beta, free while it lasts, updates as fixes and speedups land. No dead links here on purpose — if a button on this page is clickable, it works.

A macOS (Apple Silicon) build is planned after the Windows beta stabilizes. M-series Macs only; Intel Macs are not planned.

System requirements

  • OS: Windows 10 or 11, 64-bit (Apple Silicon macOS build planned)
  • Memory: 16 GB RAM
  • GPU: NVIDIA GPU recommended (CUDA + NVENC fast path) — any OpenCL-capable GPU works
  • Storage: fast local disk with room for your source clips plus the exported panorama (a full match is easily 50–100 GB end to end)
  • Footage: two overlapping recordings of the same field — works with GoPro footage (Linear FOV recommended) and similar action cameras

Stitching two 4K60 streams is heavy, but it runs on modest hardware — it just takes longer. See the benchmarks for what to expect from a machine like yours.

Step 2 · Track the ball

BallTrack AI

Takes a wide or panoramic match video and renders a broadcast-style MP4: AI ball and player detection drives a virtual camera that pans and zooms to follow the play — no operator, no PTZ rig. It’s built for FullField Stitch panoramas, but it works standalone with any wide, static-camera recording of a match.

Coming soon

Windows beta installer

The BallTrack AI beta installer is being packaged alongside the stitcher’s. Nothing to download yet — this card gets a real button when it ships.

Download for Windows — coming soon

Beta notes: a separate download from FullField Stitch — install either app or both. Free while in beta, same feedback deal.

Windows only: the AI detection runs on CUDA, so a macOS build is not planned.

System requirements

  • OS: Windows 10 or 11, 64-bit (Windows only)
  • GPU: NVIDIA GPU with CUDA required, RTX 20 series or newer only — it runs YOLO detection on every frame
  • VRAM: 12 GB minimum (RTX 3060 12 GB or better) for 4 detection workers, which is where the speed gains flatten; 16 GB is the sweet spot — all 5 workers, or a heavier model at fewer workers. 10–11 GB cards top out at 3 workers; 8 GB is below the floor. More than 16 GB does not run faster — only 24 GB has been validated in real runs
  • Memory: 16 GB RAM
  • Input: any wide, static-camera match video — a FullField Stitch panorama is ideal, but not required
  • Storage: fast local disk with room for the source video plus the rendered output

The NVIDIA requirement is hard: the CPU code path is orders of magnitude too slow to be usable on a full match, so treat CUDA as mandatory. FullField Stitch above is the more forgiving of the two.

The beta deal

Both apps share the same beta program. Small beta, real conversations. Here’s what you get and what’s expected.

What beta testers get

  • The full apps, free during the beta — every feature, no watermarks, no time limits
  • Updates throughout the beta as fixes and speedups land
  • A direct line to the developer — your reports actually get read and acted on
  • A say in what gets built next

What’s expected of you

  • Share your benchmark numbers. The apps write a small local benchmark log after each export or render — sending it along helps map real-world performance across GPUs
  • Report bugs with enough detail to reproduce: what you did, what happened, what you expected
  • Patience — it’s a beta, and some matches will find bugs before you find highlights

Want in?

A beta signup contact will appear here shortly — one signup covers both apps. Until then, check back — the first Windows wave is close.

Beta signup — opening soon