A repeatable system for producing photorealistic, form-accurate demo videos across your training programs: a custom dashboard, a proven prompt method, and hands-on coaching so you run it yourself.
You want a reliable way to produce photorealistic, form-accurate workout demonstration videos on demand, across your training programs (strength, resistance-band, bodyweight, mobility and stretch, and more). Rather than commissioning each video as a one-off, the goal is a repeatable system: a simple tool where you submit a prompt, the system generates and stitches the footage, and you receive a finished video.
This document sets out what we will build, how it works, how we will work together, and the pricing structure. The end state: you produce your own workout videos self-serve, with TX having built the tool and coached you through the first few videos so you can run it independently.
Three deliverables that leave you with a working tool, finished videos, and a method you own.
A custom dashboard built around the Higgsfield generation engine. Submit a prompt or pick a saved template, the system generates the motion clips, stitches them into one continuous video, and delivers the final file. This is the core asset you keep and reuse.
TX produces the first 3-4 videos alongside you. This is where we perfect the prompting for your exercises, lock in the form accuracy, and establish the repeatable method. You end this phase with finished videos and a proven process.
A documented prompt method and a set of tested, reusable prompt templates for your exercises, so you self-serve going forward without starting from scratch each time. After the first videos, you take the controls.
The system produces accurate demo videos for any exercise across your workout programs, reproduced with correct, injury-safe form. It is not tied to one routine. The exercise types below are illustrative examples of what the system covers, not a fixed deliverable set.
Built on Higgsfield's video engine. A prompt describes the exercise and setting, the engine produces short segments of continuous, photorealistic motion, and the system stitches them into one video. This is how a short demo becomes a full follow-along routine.
Because the video is assembled from stitched segments, there can be a slight visual transition roughly every 7-10 seconds where two segments join. We minimise this by making the first prompt as precise as possible and extending the same validated prompt for repeated motion, rather than regenerating each rep from scratch. This keeps the movement consistent across the length of the video.
Testing shows the engine renders well-known, iconic movements, for example the barbell squat, very accurately from a text prompt alone. These exercises are the low-effort case and come together quickly.
Complex or less common movements, and especially resistance-band moves, are harder: the engine does not always render equipment like a resistance band or the exact movement reliably from a text prompt alone. This applies across workout types. Our solution is a reference-guided route: we drive each harder exercise from a real reference clip of the correct movement, so the generated video follows accurate, verified form rather than relying on the text description alone. This is the method we lock in during the hands-on phase, and it is how we hit the form-accuracy bar on the trickier moves.
Presented separately so you can see exactly what each part costs and what you control. TX's engineering and coaching time is one simple hourly rate; the generation credits are loaded by you, directly.
One clear hourly rate covering both the software build (the dashboard) and the hands-on assistance producing your first 3-4 videos. You pay for the work actually done, and effort drops as the method matures and you take over.
The engine runs on Higgsfield credits. You load your own credits directly, so you stay in full control of that spend and TX never bills you for it. Raw single-pass generation sits at the low end; prompt-perfection iterations to dial in exact form add on top.
Once the tool is built and the method is locked, you generate videos on your own. Your only recurring cost is the Higgsfield credits you load (Layer 2). TX stays available on the hourly rate whenever you want extra help.
The per-minute figures are an estimate to set expectations, not a fixed price. Actual credit use depends on the exercise and the number of revision passes: more adjustments to reach exact form means more credits. You always see and control this spend directly, since the credits are yours.
Scope-first, then we begin. Simple and staged.