A teleprompter for field reporting — the piece to camera in one take

A piece to camera is thirty seconds of screen time and, on a bad day, an hour of location time. The words are not optional: names with the right spelling, the charge exactly as the court listed it, the figure the council press office actually confirmed. You wrote them in the car, someone on the desk approved them, and now you have to say them — precisely — standing in the wind, in front of the thing you are reporting on.

The deadline does not negotiate. The bulletin airs at one, the editor wants the piece by half twelve, and the light, the presser or the police cordon set the schedule long before your memory gets a say. Every fluffed name means walking the take back to the top, and every retake spends a minute that was budgeted for something else. Retakes also cost you the street: the third attempt at a piece to camera in front of a flooded shopfront gathers an audience, and the audience gets into the shot.

Most reporters filing from location are alone now. There is no camera operator to nod when you drift out of frame and no prompter operator, because there is no prompter — there is a phone on a tripod and whatever you can hold in your head. The studio has an operator and a sheet of glass; the roadside has neither.

And increasingly the desk wants two things from the same stand-up: a live hit into the station stream, and a clean recording for the packaged version. Delivering both from one pavement used to take a truck and two engineers.

The standard field technique is memorising three sentences at a time and shooting the stand-up in pieces, which is why so many location reports cut away to B-roll mid-thought — the edit is hiding the joins. Bullet points and improvisation work for colour, but not for the paragraph the lawyer saw before it left the building. The other field trick is the ear prompter — the script read once into a voice memo and played back through an earpiece — which produces the unmistakable cadence of someone repeating what they hear, half a second behind themselves.

Paper still gets used: a notebook held just under the lens, eyes dipping twice a sentence. In rain and wind it is worse than nothing, and after dark it needs a second phone held up as a torch.

For the live side, stations that can afford bonded encoders send them out; everyone else opens a streaming app that has never heard of a script and reads from the page in the other hand. The two halves of the job — saying it right and getting it out — live in different tools that know nothing about each other.

Talk2Camera puts the script on the phone that is filming, directly under the lens, and scrolls it as you speak. Stop while the lorry passes and it waits; skip the sentence that events have overtaken and it catches up. The speech recognition runs on the device, so the prompter works in a field with no bars at all — and the words of an embargoed script never leave the phone.

For the live hit, paste the RTMP URL and stream key from your station ingest — or from YouTube or Facebook, if that is where the outlet goes live — and stream from the same app the prompter is in. When the mobile connection dips, it reconnects on its own with backoff instead of dying quietly behind you while you keep talking. The stream needs signal, of course — the prompter is the part that does not. And the reconnect buys you the honest failure mode: a hit that stutters and comes back, rather than one that quietly became you talking to nobody.

Recording and streaming are separate decisions. The clean local recording keeps running whatever the stream does, and you can start or stop it without touching the broadcast — so a two-minute live cross leaves behind a full-quality file the packaged report can actually use, not a screen capture of a compressed stream.

If the shot needs the phone across the street on a tripod, a second phone or tablet becomes the prompter beside the lens. Pairing is pointing one device at the other screen; the code works once and expires after three minutes. The link between the two is direct — no router, no venue Wi-Fi, no account. Two Apple devices connect with no network at all; when pairing with an Android second screen, the Android device is the one that shows the code — scan it with the iPhone and the two are connected. And if the link drops mid-take, the script keeps scrolling rather than stopping dead under your eyes.

A flooded high street, 11:40, and the midday bulletin wants a live picture at noon plus a stand-up for the lunchtime package. The desk emails the approved script; you paste it into the app, set the phone on the tripod at the water edge, and pair the tablet as the prompter beside the lens.

At noon you go live into the station RTMP ingest and read the hit — the script following your voice, the recording running underneath it. After the cross you stop the stream, keep the camera rolling, and read the packaged stand-up once more, clean, with no presenter talking in your ear.

By 12:15 the local file is trimmed to the first word and filed. One location, one phone, both deliverables — and the wording is the wording the desk signed off, because it was on the screen the whole time rather than in your memory.