AI Tools for Short Videos and Visuals: How to Use Them, and Who Owns What You Make

AI tools can now turn a prompt, a song, a script or a recording into a short video or a picture. This guide covers five of them. For each: what it starts from, how to make a first piece, and who owns the result under its terms. None of them paid to be here, and none of the links earn us anything.

011. The five tools at a glance

Pick the tool by what you start with: a prompt, a song, a script or a recording.

All five use generative AI, which makes new pictures, sound, video or text from what you give it. Each one is described below from its own pages.

Tool	You start with	What it makes	Free to try?
Magnific (formerly Freepik)	A prompt or a picture	Pictures, video, voices and music, plus stock files	Free accounts, for personal, non-commercial use
Kaiber	A song, plus pictures or clips	Music videos cut to the beat	No free plan shown; a $5 trial for five days
Fliki	A script, a blog post or an idea	Narrated videos with stock pictures	Free plan, with a watermark
Descript	A recording	An edited video, captions and clips	Free plan, with 720p export
Dezgo	A prompt or a picture	Pictures and video clips	Free version, no account needed

022. Four questions before you make anything

Check ownership, training, who can see it, and whose face or voice is in it.

Each tool below answers these in its own terms of use, and the answers differ from tool to tool. On Magnific and Kaiber, the terms give free use fewer rights than paid use, and Fliki lists commercial rights only on its paid plans.

• Who owns it? Search the terms for “output” and “ownership”. Check whether free and paid plans get the same rights.
• Does it train on it? Search the terms and privacy policy for “train”. Look for a setting that turns it off.
• Who can see it? Check whether what you make stays private. Magnific, for example, treats inputs and outputs as confidential unless you share them publicly.
• Whose face or voice is in it? If a real person appears, or their voice is copied, get their permission. The tools’ own rules make that your job.

For example, if a clip is for a client’s paid advert, check your plan’s commercial rights before you start.

033. Magnific (formerly Freepik): pictures, clips and stock in one place

Magnific is Freepik’s new name, and one account reaches many image, video and audio models.

Magnific’s home page says “Freepik is now Magnific” and calls it the same team and the same technology. Its menu includes an AI Image Generator, a Video Generator, Text to Speech and a Music Generator. It also holds a stock library it describes as “250M+ photos, vectors, icons, and templates”.

How to start with Magnific

Settle on one picture before you turn anything into video.

In its pricing table, a few seconds of video usually costs more credits than one picture. In practice, make a few versions of one scene in the image generator. Pick the one you like, and only then animate it. A clip built on a picture you already like wastes fewer credits.

Who owns what you make on Magnific

Paying subscribers get Magnific’s rights in what they make; free accounts get personal, non-commercial use.

Its AI Products Terms, dated June 2026, assign paying subscribers all the rights Magnific may have in an output. That holds only if the subscription is active when the output is made. Free accounts get a personal licence to use outputs only for non-commercial purposes. The terms also let Magnific restrict outputs of certain tools to uses such as personal or editorial use. Music from its ElevenLabs-powered Music Generator may not go on streaming services such as Spotify, or be used commercially on its own as audio files or samples. The terms say Magnific does not guarantee an output is exclusive. They add that in some countries an output may not be protected by intellectual property rights. The terms let Magnific use your inputs and outputs for security and to improve its products and services. They say Magnific will not use them to train its own AI models. Its privacy policy, dated April 2026, adds that it will never use images or voices you upload to train its own or other providers’ AI models.

Our reading: if the work is for a business or a client, make it while a paid plan is active.

Worth knowing about Magnific

Prices depend on your region and change with offers, so read the pricing page on the day.

The pricing page lists Premium, Premium+ and Pro Starter plans, with prices that exclude VAT and local taxes. It was running a time-limited discount when we looked, so we don’t print the figures. The terms say you must have permission from anyone whose image, voice or identity is in what you upload. Stock files used for free must credit the site, and the paid plans remove that need.

044. Kaiber: a music video from your own track

Kaiber is made for musicians: you bring a song, and it builds video around it.

Kaiber’s home page says “Turn your song into a music video”. Its Beat Sync tool takes your pictures or clips and a track and returns up to 10 beat-synced videos. Its Canvas also has ready-made workflows such as Animate Image and Make Them Talk.

How to start with Kaiber

Start with one section of the song and a handful of your own pictures.

For example, take the chorus of a new track and five photos from the session. Let Beat Sync cut them to the rhythm. Using your own pictures keeps the rights question simple. Beat Sync is not included on the Starter plan. How many videos it makes at once depends on the plan: up to 2 on Creator and up to 10 on Pro.

Who owns what you make on Kaiber

You own your uploads and outputs; third-party content Kaiber adds may be used only within that output.

Kaiber’s terms say you keep ownership of your uploads and of the output made from them. Any third-party content in the output is excluded. You may use that third-party content only as part of the original output, not on its own. They add that outputs may not be unique, because the service can make the same or similar output for others. Kaiber says it will not use your content to train generative AI models without your permission. The same terms let it use “usage data”, which includes text you type, in de-identified or aggregated form to train and improve its products. Its privacy policy, last revised in October 2023, does not mention training. Where Kaiber offers a free tier, the terms limit it to non-commercial use.

Worth knowing about Kaiber

The pricing page shows paid plans and a paid trial, not a free plan.

It lists Starter at $10, Creator at $29 and Pro at $99 a month, with 20% off for yearly billing. A trial costs $5 for five days, then becomes the Creator plan. Before you use Make Them Talk on a photo of a real person, ask that person. Kaiber’s terms bar using the platform in a way that violates any person’s rights.

For videos that move with the sound itself rather than AI scenes, see our music visualizers guide.

055. Fliki: from a script or blog post to a narrated video

Fliki turns text into a narrated video, adding an AI voice, pictures, music and subtitles.

Fliki says you can paste a script, a blog post or a one-line prompt. It then writes the script, picks visuals and adds the voiceover, music and subtitles. It describes its voice library as “2,000+ lifelike AI voices in 80+ languages and dialects”.

How to start with Fliki

Start from text you already wrote, then fix the scenes it picks.

For example, paste a short blog post and let it build a first draft. Then swap any stock picture that doesn’t match its sentence. Listen through once before you export, so a name read oddly doesn’t slip out.

Who owns what you make on Fliki

You keep your rights, and you also give Fliki a broad licence to use your content.

Fliki’s terms say you retain all of your rights to content you create on the service. They also grant Fliki a licence to use that content to operate, improve and promote its services and to develop new products. Its privacy policy says photos or videos of faces that you upload are not used to train AI models. The policy does not say whether other content you make is used for training.

Voice cloning on Fliki

Fliki can clone a voice, and asks you to clone only yours or one you may use.

Its voice-cloning page says to clone only a voice you own or have explicit permission to use. It says Fliki requires a consent step before a clone is made. It says voice cloning comes with the paid plans. Its terms also name impersonating people without their consent as a misuse. For example, clone your own voice to narrate your channel, but never a friend’s without asking first.

Worth knowing about Fliki

The free plan is for trying it out: few credits, lower resolution and a watermark.

The free plan gives 3 credits a month, 720p videos and a Fliki watermark. Commercial rights appear in the paid plans’ lists, not the free plan’s. The page shows prices in your local currency and changes them with sales, so we don’t print them.

066. Descript: edit a video by editing its words

Descript turns your recording into text, and deleting words from the text deletes them from the video.

Our guide to AI meeting-notes tools covers its transcripts; this section looks only at making videos. Descript says you edit your words to edit your video, and can cut filler words and retakes in a click. It can add captions, and its menu includes a Create Clips tool.

How to start with Descript

Record or upload one talk, then cut it down by editing the text.

For example, film a five-minute explanation, delete the stumbles in the transcript, and pull out one minute as a short. Watch the cut once at full speed to catch any jumps where words were removed.

Who owns what you make on Descript

You own your work where the law protects it; Descript may train on it unless you opt out.

Descript’s terms say that, between you and Descript, you own your input and output, to the extent the law protects them. They say Descript may use inputs and outputs to train its own models unless you have opted out. Its privacy policy says you can stop your projects being used to improve the service by turning off the “Share Data with Descript” setting. The terms add that its agreements with outside AI providers bar them from training on your inputs and outputs.

To make an AI voice of someone, that person must give permission, and submitting another person’s voice recordings without authorisation is prohibited.

Worth knowing about Descript

There is a free plan, and the paid plans are priced per person.

The free plan includes 60 media minutes a month and 720p export. On yearly billing, paid plans run from Hobbyist at $16 to Business at $50 per person a month. Monthly billing runs from $24 to $65.

077. Dezgo: try image models without an account

Dezgo lets you try many image and video models in a browser, and its free version needs no account.

Dezgo’s FAQ says there is a free version you can use without registering. The free version has limits such as rate limits and lower image resolution, and some powerful video models are not in it. Its image tools include background removal, upscaling and editing an image from text.

How to start with Dezgo

Run one prompt through two or three models and compare.

For example, describe one product shot in a sentence, generate it on two different models, and keep the clearer result. Note which model made it, because the licence that applies depends on the model.

Who owns what you make on Dezgo

Dezgo says you may use outputs commercially, but each model’s own licence still applies.

Its FAQ says you own the rights to use each output as you wish, including commercially. The condition is that you follow each AI model’s licence terms. It says this applies to both the free version and the paid Power Mode. Its licences page lists the Apache License 2.0, the CreativeML Open RAIL-M licence and Google’s Gemini API Additional Terms of Service.

Our reading: before you sell anything made there, read the licence for the model you used.

Worth knowing about Dezgo

It is ad-supported, and its public pages give no price for the paid version.

Dezgo’s privacy policy says it uses third-party advertising companies to serve ads. We found no price for Power Mode on its public pages. Its privacy policy does not say whether it trains on what you make. Its terms ban sexual content, and say you need the rights and permissions to use any image of a real person. The company behind it is Dezgo SAS, registered in Paris.

088. Who owns AI-made work, beyond the tools’ terms

Terms hand over a tool’s rights, but Magnific warns some countries’ laws may not protect an output.

Descript’s terms, for example, give you ownership only “to the extent protectable under applicable law”. In the US, the Copyright Office’s January 2025 report says copyright “does not extend to purely AI-generated material”. It says that with current generally available technology, prompts alone do not give enough control. Using AI to assist rather than stand in for human creativity does not affect copyright in the output, it adds.

The report is about US law only. Two of the five tools also say that another user may get the same or a similar result. If a project depends on owning what you made, ask a lawyer where you live.

099. Faces, voices and labels

Use only faces and voices you have permission for, and label realistic AI content when a platform requires it.

A video, photo or audio clip that seems real but has been manipulated with AI is what people call a deepfake. YouTube requires creators to disclose AI content when it appears realistic. Its must-disclose list includes content that makes a real person appear to say or do something they didn’t do. It also says creators don’t need to disclose non-realistic content that’s made with AI. TikTok requires creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people.

Magnific’s terms tell users to disclose clearly when content could be taken for a deepfake, citing Article 50(4) of the EU’s AI Act. For example, a cartoon fox reading a recipe is plainly not real. A lifelike clip of a named chef reading it could pass for real.

If your own face is the worry, our YouTube 2026 guide explains YouTube’s likeness detection.

1010. Where to go next

For a whole first video, and for other tools, these guides go further.

• How to make 3D-style motion graphics videos with Claude and AI: the whole method, from script to edit, with prompts.
• Google Flow explained: Google’s own video tool, from its help pages.
• Music visualizers compared: tools that turn audio into moving visuals.
• AI tools that turn meetings into notes: Descript’s transcripts, and what note-takers do with your recordings.
• AI side hustles: the rules that apply when you sell work made with AI.

11Quick answers

12Use this

13Quick quiz

How we checked this · Last checked October 2026

• We read each tool's own home, pricing, terms and privacy pages in October 2026. Kaiber's pages, and the price toggles, were read in a web browser because they load by script.
• We did not sign up for or pay for any tool, so features behind a login were not tested.
• Prices change with region and offers. We print them only where a page showed a clear figure in US dollars.
• We looked at more tools than are listed here, and include only those whose owner and terms we could read.
• The Copyright Office report covers US law only. We did not read other countries' laws, or the text of the EU AI Act, which appears here only as Magnific's own citation.

How we check every article →

14Sources