How to Add Voiceover to a Video with AI, Step by Step

A video voiceover is narration spoken over picture: a separate audio track you add to the footage, rather than a performance delivered on camera. If you want to know how to add voiceover to a video, you can record your own voice or generate an AI voiceover from a script with text-to-speech, then place that track on a timeline and mix it under the picture.
This guide teaches the text-to-speech path: write lines that sound spoken, pick a catalog voice, generate and preview the audio, download it, and finish the sync in an editor you already use. CinemaDrop can generate per-shot speech and hand you the assets; it does not compile a finished video file.
- A voiceover sits over the picture, not as an on-camera performance and not as a dub into another language.
- Write the narration for speech, generate or record the track, download the audio, place it on a timeline, and duck any music under the voice.
- In CinemaDrop you can generate per-shot speech with ElevenLabs text-to-speech (Lite and up) and, on Basic and up, export a ZIP of storyboard assets.
What is a video voiceover?
A voiceover is spoken narration that comments on, explains, or carries a story while the picture does something else. Documentaries use it to steer attention, explainers to walk through a process, and short films when thoughts or a narrator sit over images. Faceless videos use it when no one appears on camera.
The words have to fit the shots, the voice has to be clear, and the track has to leave room for music and effects. Learning how to add voiceover to a video starts with that distinction: it is not dialogue recorded on set, and it is not the same as making a face appear to speak.
We’ll use one short, original scene as a through-line: a fictional harbor lamp-keeper, Arun, rows out at dusk to light the channel markers before the night ferry. The picture shows the crossing; a narrator, not Arun on camera, tells us what is at stake.
Voiceover vs dubbing vs lip sync vs speech-to-speech
These terms get used interchangeably in search results. For planning a film, keep them separate.
| Job | What changes | What CinemaDrop can and cannot do |
|---|---|---|
| Voiceover | You add a narration track over picture. Faces in the frame are not being re-voiced to new mouth shapes. | You can generate per-shot speech with text-to-speech (Lite and up) and download the audio. CinemaDrop does not compile a finished video with that track mixed in. |
| Dubbing | You replace spoken language in a finished cut, usually for another market, and adapt the lines so they fit the picture. | CinemaDrop has no translation or dubbing workflow. Use our guide to AI dubbing for that job; this post does not retell it. |
| Lip sync | You align a mouth, on a still or in existing footage, to a speech track so the face appears to say the line. | Lip sync is available on Basic and up; this article does not rate quality. See AI lip sync for filmmakers, or talking photo AI if you are starting from a still. |
| Speech-to-speech | You upload a performance you are allowed to use and convert it into a catalog voice, keeping the timing and delivery of the original take. | Available on Plus and up, as conversion into an ElevenLabs catalog voice, not custom voice cloning or multilingual dubbing. See the speech-to-speech guide. |
CinemaDrop does not offer custom voice cloning. If you need a face to deliver the line on screen, that is lip sync or a talking photo, not a voiceover.
Should you record your own voice or generate one?
People looking for how to do a voiceover for a video often mean a live recording: a quiet room, a microphone, a written script, and a test take. That path still works. If the film needs your own cadence, an accent you already have, or a performance you can adjust while you watch the picture, record it.
A short checklist is enough for most first takes:
- Write the script first, then read it aloud once before you open a recorder.
- Use a quiet room. Turn off fans and notifications. Keep a consistent distance from the mic.
- Record a test sentence, listen on headphones, and fix plosives, rustle, or room echo before the full pass.
- Leave a little silence at the head and tail of each take so you can trim.
The rest of this post teaches the other path: generating an AI voiceover for video from text. That fits when you want a catalog voice, you are iterating on wording, or you are not the right speaker for the film. You can still record a scratch track and replace it later, or convert an authorized recording into a catalog voice on Plus and up.
Write the narration so it sounds spoken
Text that reads well on a page often sounds stiff when it is spoken. Write the way you would tell the story out loud, then generate.
- Keep sentences short. If you have to take a breath in the middle, split the line.
- Treat punctuation as pace: a period is a stop, a comma is a smaller pause, and an em dash can hold a beat.
- Map each line to a shot or a timestamp before you generate, so you are not writing one paragraph to cover five images.
- Read the lines aloud. If you stumble, fix the wording first.
For the harbor scene, three lines might cover three shots:
- Opening image: “An hour before the ferry, the channel is only a dark line between two rows of unlit markers.”
- The crossing: “Arun has made this crossing every evening for twelve years. Tonight the wind is the kind that hides the bell.”
- The last lantern: “If the last lantern fails, the ferry waits until morning.”
Those lines do not describe what the picture already shows. They add time, habit, and a consequence. If a sentence only repeats the image, cut it.
How to add an AI voiceover to a video, step by step
Here is how to add voiceover to a video with a generated track. CinemaDrop can do the generate-and-download step inside a storyboard; an external editor does the rest.
- Write the narration and attach each line to a shot or a time range, as in the harbor example above.
- Choose a catalog voice that fits the scene and listen to samples before you generate the full line. You are picking a library voice, not building a custom model of a real person.
- Generate the speech and preview it without picture first. Listen for missed names, odd stress, or a pause in the wrong place.
- Fix the wording or the pronunciation in the text, then regenerate. Change one thing at a time so you can hear what improved.
- Download the audio for each line or shot.
- Place the files on a timeline in an external editor and nudge the in-points so each line lands with its shot.
- Duck music under the voice, add any effects, and do a final listen on speakers and on a phone.
If you are starting from audio and want to build picture afterward, that is the opposite job. CinemaDrop’s audio-to-video path is for that direction, not for placing a new voice over finished shots.

Adding narration in a storyboard, then finishing the mix
In a CinemaDrop storyboard, speech lives on the shot. Open Speech and follow the documented text-to-speech walk-through: choose a voice with Select, Sample, and Use Voice, then open the advanced menu to set stability (Creative, Natural, or Robust) or type the line there. You can mark emotion in [brackets], for example [quietly]. Click Generate Speech, play the take, download it, or regenerate after you edit the wording. The latest take is attached to the shot.
The engine is ElevenLabs text-to-speech v3, with 50+ catalog voices and filters such as age, gender, and use case. It costs 1 credit per 10 characters, rounded up, and is available on Lite and up. The speech field in the UI accepts up to 5,000 characters, though shorter, per-shot blocks are usually easier to time. Preview every take yourself; this article does not claim a particular sound, language coverage, or turnaround time.
If you already have a performance you are allowed to use and want it spoken in a catalog voice, switch to the Speech to Speech tab on Plus and up, upload the file, pick the voice, and generate. That conversion costs 1 credit per second of input audio. It is not a clone of the original speaker, and it is not a substitute for dubbing.
When the shots are ready, Basic and higher plans can export a ZIP of the storyboard assets. Lite includes text-to-speech (50 one-time credits, watermark, personal use) but not video generation and not that ZIP export, so a Lite user would download the speech from the shot and finish elsewhere. CinemaDrop is not an NLE: there is no timeline, and there is no compiled MP4 with the voiceover baked in. Bring the picture, the speech files, and any music or effects into the editor you already use—Premiere, DaVinci Resolve, CapCut, iMovie, Clipchamp, Camtasia, or another tool—and mix there. Plans and rates are on the pricing page.

Captions, mix and a last listen
Once the voice sits on the timeline, do the last pass as if you were in a screening.
- Add captions or subtitles that match the final wording, including any line you changed after generating.
- Lower music when the voice is speaking, then let it rise in the gaps. Leave a little air at the head of a sentence.
- If the scene needs atmosphere under the narration, place effects on a separate layer. Our guide to an AI sound effects generator covers that job.
- Listen once on decent speakers and once on a phone. If a name or last word is unclear on the small speaker, fix the wording or the level and listen again.
Rights, consent and what “clone speech” actually means
Use footage, music, and performances you have the right to use. A generated voice does not give you permission to narrate over someone else’s film, and it does not clear a song.
Do not try to recreate a real person’s voice from samples in order to pass the track off as them. CinemaDrop does not offer custom voice cloning. A homepage line that says “clone speech” refers to speech-to-speech: converting an authorized recording into a catalog voice, not building a model of someone’s identity.
When you publish, say so if the narration is synthetic. Platform rules and local law differ, and this is not legal advice. See our guide to AI copyright for a broader rights overview.
FAQ
How do I add voiceover to a video?
Write or record a narration track, then place it on a timeline against the picture and mix it. If you are generating the track, use the step-by-step above.
How can I add audio over a video?
Import the file into an editor, line it up with the picture, and set levels, whether the audio is a voiceover, music, or effects. “Insert voice over in a video” is the same job.
Can ChatGPT do voice overs?
ChatGPT is useful for drafting a narration script. ChatGPT Voice, on OpenAI’s help pages, is a conversational mode, not a tool that places a voice track onto a video. CinemaDrop’s text-to-speech engine is ElevenLabs, not ChatGPT.
How to do voiceover in CapCut?
You can record a take in CapCut or import a text-to-speech file, then place it on the timeline and mix. This post is not a CapCut tutorial; the same hand-off works in other editors.
What are some free voice over apps for videos?
This is not a ranked app list. CinemaDrop’s free Lite plan includes text-to-speech (50 one-time credits, watermark, personal use) but no video generation and no ZIP export of a storyboard. It does not produce a free finished video with the voice already mixed in.
Which AI voice generator is the best?
There is no single best voice. Pick a catalog voice that fits the scene and preview it on your lines. CinemaDrop uses ElevenLabs text-to-speech for that step.
Can you add a voice over to a video on iPhone or on a Mac?
Yes. You can record in an app such as iMovie, or import a text-to-speech file and place it on the timeline. This guide is not a device walkthrough.
How do I add voice over in Camtasia?
Camtasia can record a voiceover or accept a text-to-speech file you import. Use its documentation for the editor controls; the script-and-generate steps above still apply first.
Generate the speech, then finish the mix
The durable method is a script written for speech, a catalog voice you have previewed, a downloaded track, and a last listen in an editor. Start a storyboard in CinemaDrop, generate per-shot speech, export the assets you have access to, and finish the mix in your editor. Plans are on the pricing page. If the next step is a faceless episode, see our guide to starting a faceless YouTube channel with AI.


