Almost every short-form video now opens with two hooks at once: a line of text burned into the frame and a spoken sentence underneath it. Most of the time they say the same thing, which wastes one of them.
They arrive at different speeds
On-screen text is available instantly and in full. The viewer takes it in as a shape before they read it as language, and they finish reading it well before a spoken sentence of the same length has finished. Speech arrives serially, one word at a time, and cannot be skimmed.
That difference in speed decides what each is good for. Text is the right channel for the promise — the thing that should be known immediately and can be absorbed at a glance. Speech is the right channel for the turn — the thing that lands better when it arrives in sequence, with timing and tone doing part of the work.
The cancellation problem
When both channels carry the identical sentence, the spoken version is redundant by the time it is half delivered. The viewer has already read it. The first two seconds of audio are then spent confirming something known, which is precisely the stretch that should be creating momentum.
It is worse than neutral, because it teaches the viewer that the audio is not load-bearing. A viewer who learns that the speech only repeats the caption has been given permission to stop listening.
If the text and the voice say the same sentence, you have two hooks doing one hook's work.
What to do instead
Split the job. Let the text state the promise plainly and let the speech start somewhere the text did not go — the complication, the exception, the reason the promise is not obvious. The two then compound: the viewer holds the promise in view while the voice argues with it.
A useful constraint is that the spoken opening should not be a readable summary of the text. If it could be swapped for the text without loss, it is not doing separate work.
The muted-playback caveat
Text has one structural advantage: it survives playback with no sound. That is a real consideration, and it is the honest argument for putting the promise in text rather than trusting it to audio.
But it is not an argument for filling the frame. Instagram has said plainly that it aims to make reels that are majority text less visible, alongside low-resolution, watermarked, muted and bordered reels. A text hook is a line, not a wall. The format that hedges hardest against muted playback is also the format the platform demotes.
Placement
Two practical constraints, both boring and both frequently ignored. Text placed too low collides with the caption and the interface controls, and text placed too high collides with the top-edge furniture. The safe middle band is narrower than most editing previews suggest, and text that is technically on screen but sitting under a button has failed in the only way that matters.
The second: if the text is going to change during the opening, it should change after the promise has been read, not during. Motion in the first moment competes with reading, and reading loses.
See also what a hook actually has to do, and where captions help and where they get in the way.