How to Make a Video With Images and Voice-Over, Step by Step

To make a video with images and voice-over, you first write the script, record the voice-over (your own or an AI one), split the text into short scenes, put one image in each scene for exactly the length of its piece of voice and give it movement. Then you add the subtitles and export as a horizontal MP4. The order matters: if the voice goes first, the images adjust to it and not the other way round.

Script first, written for the ear

A video of images and voice rests on the narration: if the text cannot be understood when heard, no image will fix it. Write it so that it sounds good aloud, with short sentences and one idea per sentence. Read it aloud before recording and cut whatever makes you stumble.

As you write it, think in images already. Every two or three sentences, what is seen changes, so the text should contain things that can be shown: a place, an object, a gesture, a fact. Long abstract sentences (“the importance of innovation in the current context”) are hard to illustrate. If it is your first script, the process is in how to write a script for YouTube.

The voice-over: your voice or an AI one

There are two routes and both work:

• Recording your voice. It gives the channel personality. With a decent microphone, a room without echo (curtains, furniture and rugs help) and the text in front of you, you get a clean result. Record in blocks or sentence by sentence: it is easier to repeat a sentence than a paragraph.
• An AI-generated voice. You paste the script into a text-to-speech tool, choose the voice and download the audio. It is quick and consistent, but listen to the whole result: proper names, acronyms and figures are where it stumbles most.

If you are torn between the two, AI voice-over or your own voice gives you the criteria to decide. Whichever you choose, save the voice scene by scene or mark where each one starts: it will save you work when editing.

Splitting into scenes and choosing an image for each

A scene is a piece of narration with a single image. A practical reference: sentences grouped up to around twenty words, which at a calm narration pace is about eight or nine seconds. Any shorter and it feels frantic; much longer and the image stays still for too long.

For each scene, note what is seen. For example, in a video about a traditional recipe:

Narration	Image	Movement
“The recipe reached the village with the shepherds who came down from the mountain in winter.”	Shepherds with their flock coming down a snowy track.	Sideways pan
“It was made in a single clay pot, over the fire in the hearth.”	Clay pot on the embers of a fireplace.	Slow zoom-in
“Three ingredients: potato, paprika and patience.”	On-screen text with the three ingredients.	Static

The images can be your own photos, licensed assets or AI-generated images. If they come from outside, check their licence and note it down (copyright-free photos for videos explains how). If you generate them, make sure they all share a style.

Syncing: the voice sets the timing

The rule that saves the most time: each image lasts exactly as long as its piece of voice. If you edit the other way round (first the images with a fixed length and then the voice on top), you will end up stretching and trimming everything.

1. Place the whole voice track on your editor's timeline, or the audio of each scene one after another.
2. Mark where each scene starts on the audio waveform: the pauses between sentences show up as flat stretches.
3. Put each image from one cut to the next. That way the image changes just when what is being told changes.
4. Let the endings breathe. Half a second of margin at the end of a section helps the viewer take it in before moving on to the next.

If you record scene by scene, everything is easier: each audio file brings its own length and you only have to put the image in with that same length.

Movement, subtitles and export

Movement. An image that stays still for eight seconds is noticeable. A slow zoom-in or a gentle pan makes it pass for a camera shot; we explain how in the Ken Burns effect. Keep the static shot for on-screen text and graphics that have to be read.

Subtitles. Since the script is already written, you have the exact text of what is said. YouTube lets you supply that text and syncs it with the video: in Subtitles, you choose Auto-sync and enter the video's text or upload a transcript file, which must be in a language supported by YouTube's speech recognition technology and in the same language spoken in the video.

Export. YouTube's recommended upload settings are an MP4 with H.264 video and AAC-LC, Opus or Eclipsa Audio at 48 kHz. YouTube's standard aspect ratio on computers is 16:9, so export in landscape unless you are making Shorts. And do not change the frame rate when exporting: YouTube asks you to encode and upload at the same frame rate the video was recorded at, and cites 24, 25, 30, 48, 50 and 60 frames per second as the most common.

How the Estudio de Canales does it

The Estudio de Canales works best with narrated videos with images, and follows this same order:

• Scenes from the approved script. Splitting into scenes does not use AI: the narration of the scenes is, word for word, the approved script. It groups sentences up to about 22 words per scene.
• The voice, scene by scene. AI voice, choosing between several models and narrators, or your own voice: you upload the recording of each scene (WAV, MP3, M4A, OGG or WEBM) or record it in the browser with a teleprompter.
• The voice sets the timing. When a scene's voice arrives, its duration becomes that of the real audio, not the one estimated from the word count.
• One image per scene, generated with AI or your own, with its licence and source noted, and a movement: Ken Burns, pan, zoom or static.
• Cloud editing with your approval. A person approves each scene; then a preview at 1280×720 is produced and, when you approve it, the master at 1920×1080 and 30 frames per second, as an MP4 with H.264 and stereo AAC audio at 48 kHz.
• Subtitles from the script. The package to upload includes an .srt built from the approved script and the real audio duration of each scene.

What it does not do: it does not add background music (the video's audio is the voice of the scenes), it does not add transitions between scenes (it goes from one to the next with a straight cut), it does not generate AI video clips and it does not publish the video: you download it in the package and upload it yourself in YouTube Studio. Today it produces horizontal 16:9 video, not Shorts.

Sources consulted

Checked on 11 October 2026. We link to the original source so that you can consult the current text.

• Añadir subtítulos. Google · Ayuda de YouTube.
• Configuración de codificación de subidas recomendada. Google · Ayuda de YouTube.