Speak the prompt instead of typing it

Both hands were on the lamp. Typing a prompt would have meant putting it down, losing the angle, and starting the photo again. Speaking was the practical path: the still was already in frame, and the instruction was one sentence about how it should move. A spoken prompt is not a different feature with different rules. It is the same line, heard instead of typed.

On photo to video, that line still has to name the motion and what must stay. Standard, Ultra, and Express all take it. The photo remains the subject. The AI photo to video homepage is the studio, whether the sentence arrived by keyboard or by voice.

When speaking is the better way in

Use it when you are holding the product, walking through a shop, or capturing in the Android app and you do not want to set the phone down to type. Use typing when you are at a desk and you want to edit a word without recording the sentence again. The generation does not score one higher than the other. A bad spoken line and a bad typed line fail the same way: they ask for a new product.

Say less than you think

People ramble when they talk and edit when they type. The ramble is how a lamp prompt becomes a room tour. Say one motion. Then say what must stay. For the lamp: slow push-in, keep the lamp, the shade, and the color as shown, camera stays still, no new objects. Stop. If you hear yourself adding a second scene, that scene is another clip.

Look at the line the app wrote down. A misheard word is still the prompt you are about to spend credits on. Fix it before you generate. This is the step that makes speech as safe as typing.

Then it is the same studio

Express if you are still in the room and you want to see the hook before you pack the product away. Ultra if the line is right and the object has to look clean. Standard if that pass fits. The clip lands in the gallery on the same account. On the web you can add the logo, combine a second shot, or publish. You do not retype the prompt into a different project to finish the file.

Speak, check, generate

1. Frame the product first The photo is the subject. Get the still before you talk.
2. Say one motion What moves, what must stay, no new objects. Leave the second beat for another clip.
3. Read the line Correct a misheard word before you generate.
4. Pick the pass Express to check the hook in the room. Ultra for the clean version. Standard when that pass fits.
5. Finish on the same account Brand, download, or publish from the gallery. The spoken prompt does not need to be spoken again.

The lamp, said once

The frame was already good: shade inside the photo, cord leaving the bottom edge, wall plain. They spoke while still holding the lamp. The first take rambled into the room, the evening, and a person sitting down. They stopped, looked at the text, and erased it. The second take was one breath. Slow push-in, keep the lamp, the shade, and the color as shown, camera stays still, no new objects. The text matched what they had meant. They generated Express before they set the lamp on the floor, saw the shade survive, and left the clean Ultra pass for the desk.

At the desk nobody re-recorded the sentence. The line was already on the project. The logo went on the wall side, which the push-in never crossed. A misheard earlier test, from a different afternoon, had generated “slow push and spin” because they had not read the line. That clip rotated the shade into a shape the photo did not contain. Reading the words is the entire safety check speech needs.

Speech does not get a looser standard

Because talking is easier than typing, the prompt gets longer unless you decide it will not. One motion. What must stay. No second scene. If you hear a second scene, that is another photograph or a next scene after this clip exists. Standard, Ultra, and Express do not listen more carefully than they read. They render the line you left on the screen.

Finish on the same account. The point of speaking was to capture the still while your hands were full, not to create a phone-only video you will describe again later. Brand it, download it, and keep it in the gallery.

If the clip replaces the lamp, the sentence was too long or the line was wrong. Shorten it, read it, and generate again from the same photo. Do not switch passes and hope the extra words behave.

What the phone actually heard

Both hands were on the lamp, so typing would have meant putting it down and losing the angle. Speaking was the way in, not a different studio. The phone wrote the line down. They read it before they generated, because speech is only the input. The sentence that spends the credits is the one on the screen. The first attempt heard “gold shade” when the shade was linen. That single wrong word was a request to repaint the product. They fixed the line with a thumb, then generated. Express was enough in the room, since they needed to know the push-in held before the angle was gone. Ultra waited until the line was right and the shade had to look clean on the page. Standard was available for a shot where that pass fit.

After the words were right, nothing else about the studio changed. The same account, the same balance, the same gallery, the same three passes, the same logo placed later on the web where the frame was easier to see. A spoken prompt does not get a looser standard and it does not get a hidden mode. If the clip replaces the lamp, the sentence was too long or the line was wrong. You do not switch passes and hope the extra words start behaving.

Say less than a typed paragraph, not more. One motion. What must stay. No objects you did not photograph. Photo to video is the path whether the line was spoken or typed. The homepage is the studio you open on the desk to finish the file the phone already started.

The lamp stayed the lamp when the line stayed boring. They branded that clip, downloaded the MP4, and left the spoken draft in the gallery. Next time the hands are full, the method is the same: frame the product, say one motion, read the words, then spend the generation.

The second take was quieter. They said the push-in, said the lamp had to stay the lamp, and stopped talking. The phone’s transcript matched the photograph, including the linen shade, so they generated. Talking through a whole commercial while the product is in your hands feels productive and writes a prompt the still cannot defend. The extra clauses arrive as new objects, new rooms, and a logo you should have placed later with your eyes on the frame. Silence after one motion is part of the method. So is the glance at the text. A spoken workflow that skips the reading step is just a faster way to spend a generation on a misheard word.

On the desk the clip was ordinary in the way they wanted. The lamp matched. The logo found a corner. The spoken sentence was still in the project if they needed to see what they had asked for. Speech did not create a second kind of credit or a second kind of pass. It created the line. Everything after that line is the same studio you would have used if you had typed it slowly at a keyboard.

Later they typed the same line on the laptop for a second lamp and got the same kind of clip. Speech had not been magic. It had been a way to keep holding the first lamp. The standard stayed the standard: one motion, the object as photographed, the words read before the credits move. If the transcript and the product disagree, the transcript loses.

They kept the transcript next to the photograph in the project notes for the lamps that followed. When a new spoken line grew a room, a hand, or a second product, the old transcript showed how little had been required the day the shade stayed linen. The correction was always in the text, before the pass, before the credits. Speech is fast. The reading step is what keeps it honest.

Read the line, then spend the generation. The lamp stayed linen because that reading happened while the product was still in the frame.

This job is covered in the product-still guide.

Where this fits

This guide supports the AI photo to video studio on the KuvaVid homepage. The page for this job is photo to video.

Keep reading

KuvaVid · All articles