Teach Pendant vs. Video Demonstration: Two Ways to Program a Robot
A clear-eyed comparison of teach pendant programming versus video demonstration for cobots. What each method requires, where each works best, and why the gap matters for manufacturers.
Teach Pendant vs. Video Demonstration: Two Ways to Program a Robot
There are two broad camps in robot programming today: traditional teach-pendant methods, and newer demonstration-based approaches using video and language. Understanding what each actually requires — and where each breaks down — is essential before you commit to either.
What Is a Teach Pendant?
A teach pendant is the handheld controller that comes with almost every industrial robot arm. It is a ruggedised touchscreen or keypad that lets an operator manually jog the robot arm to specific positions and record those positions as waypoints in a program.
A typical teach-pendant workflow looks like this:
- Put the robot in "teach mode" (usually requires the safety key)
- Jog the arm to position 1 using directional buttons or a joystick
- Record the position as waypoint P[1]
- Define the motion type (linear, joint, circular) and speed
- Repeat for every waypoint in the task
- Write conditional logic and signal I/O in the manufacturer's language (URScript for UR, INFORM for Yaskawa, RAPID for ABB)
- Test the program in reduced-speed mode
- Commission at full speed with safety present
For a 10-waypoint task, this typically takes a trained specialist 4–8 hours on the teach pendant alone, before commissioning.
What Teach Pendant Programming Requires
Specialist Training
Each robot brand has its own pendant UI and programming language. A programmer fluent in UR's PolyScope is not automatically fluent in FANUC's ROBOGUIDE or Yaskawa's INFORM. Cross-brand skill takes years to develop.
Most manufacturers without in-house robotics staff must hire an integrator for any teach-pendant work — which is why integrator costs dominate robot programming budgets.
Precision Work in a Production Environment
Jogging a robot arm to precise waypoints in a factory environment requires concentration, patience, and experience. The process is inherently iterative: jog, record, test, adjust, repeat. Doing this accurately on a factory floor with noise and activity around you is harder than it sounds.
Documentation Discipline
A teach-pendant program is only as useful as its documentation. If the waypoint at P[7] represents "approach position above jig," that needs to be written down somewhere — because the pendant itself rarely captures intent, only coordinates. Programs without good documentation become opaque quickly.
What Is Video Demonstration?
Video demonstration approaches (sometimes called "learning from demonstration" or LfD in the research literature) capture a human performing the target task on video, then use AI to extract the motion structure and translate it into a robot program.
The workflow looks very different:
- A worker demonstrates the task naturally (or near-naturally) in front of a camera
- The worker describes what they are doing in plain language: "pick up the red part, place it in the jig, push until it clicks"
- The AI system segments the video into discrete steps, infers positions and forces, and generates a robot trajectory
- The result is reviewed in a 3D simulation before touching the hardware
- The program is deployed to the robot with one click
The critical difference: the worker performing the demonstration does not need any robotics training. They just need to be able to do the task themselves and explain it out loud.
Where Teach Pendant Programming Works Well
Teach pendant methods are mature, well-understood, and reliable. They work well when:
- The task is stable and rarely changes. A task programmed once that runs for years amortises the programming cost.
- Extreme precision is required. For micron-level positioning (electronics assembly, laser welding), a trained specialist with a pendant and force-torque sensor calibration will typically outperform AI-generated trajectories today.
- The manufacturer has in-house robotics expertise. Large automotive manufacturers often have internal robotics engineers who are fluent in pendant programming — for them, the cost equation is different.
- Complex custom motion paths. Spiral welds, complex surface following, and other geometric tasks are still better handled by specialists with simulation tools than by demonstration.
Where Teach Pendant Programming Breaks Down
For the majority of small and mid-size manufacturers, teach pendant programming creates friction that kills automation projects before they start:
- Task change frequency. Job shops and high-mix manufacturers change robot tasks constantly. Each change requires the integrator back — pay again, wait weeks.
- Specialist scarcity. In most regions, there are not enough qualified robot programmers to serve demand. Wait times for experienced integrators are measured in months.
- Knowledge concentration. When one engineer holds all the robot knowledge, their departure is a business risk.
- Cost per task. For manufacturers with many small tasks to automate, paying €5,000–15,000 per task in programming costs makes the maths unworkable.
Where Video Demonstration Works Well
Video demonstration approaches are strongest when:
- Workers, not engineers, need to own the robot. If the factory needs the robot operator to be able to set up new tasks, demonstration is the only practical path.
- Tasks change frequently. High-mix environments where task changeover is measured in hours or days see the biggest benefit.
- Speed to first task matters. Getting a robot running in hours rather than weeks can be the difference between a project going ahead and being shelved.
- Pneumatic grippers are involved. Demonstration-based systems can capture the pragmatic details of pneumatic gripper sequences (open, close at position, verify grip) that are tedious to specify manually on a pendant.
Where Video Demonstration Is Not the Right Answer Yet
To be clear-eyed: demonstration approaches have real limitations in 2026:
- Sub-millimetre precision tasks. Where tolerances are under 0.1mm, specialist programming with calibrated tooling still wins.
- Highly complex motion paths. Welding along curved surfaces, complex interpolation tasks, or paths with many constrained degrees of freedom are not well suited to demonstration today.
- Novel gripper configurations. Custom end-effectors that the system has not been trained on require more configuration.
No honest vendor will claim otherwise. Understanding these limits helps you make the right choice for each task.
The Practical Decision Framework
For a given task, ask three questions:
1. How often will this task change? If the answer is "rarely or never," either method can work. If "monthly or more," demonstration-based programming pays back quickly.
2. Do you have in-house robotics expertise? If yes, teach pendant may be faster for your first few tasks. If no, the specialist cost of teach pendant work will dominate the decision.
3. How many tasks do you need to automate? For one task, the programming model matters less. For 10+ tasks across a year, the total cost difference between €5,000–15,000 per task (traditional) and a flat subscription is substantial.
The Bottom Line
Teach pendant programming is not going away — it is the right tool for high-precision, rarely-changing tasks in facilities with in-house robotics expertise.
For everyone else — small and mid-size manufacturers with frequent task changes, no specialist roboticists on staff, and pressure to get more robots working faster — video demonstration represents a genuine step change. The question is not which method is "better" in the abstract. The question is which method fits your production reality.
See also: