Skip to main content
THeAILAND.com
ไทย

Search content

Using AI to Create Video: What It Can Really Do, How Good It Is, and the Copyright Questions

Use-case ~11 min Updated 18 มิถุนายน 2569

AI for Real Work AB113

Using AI to Create Video: What It Can Really Do, How Good It Is, and the Copyright Questions

For content teams, marketers, and business owners weighing whether to bring AI video into a real production pipeline

Over the past year, typing a short prompt and getting a video clip back within a few minutes has become something you can actually do at a level usable for real work, not a toy in a research lab anymore. For teams that have to feed content onto social media every day, or businesses with limited production budgets, this is a tool worth watching. But the question executives and team leads should ask is not “is it cool?” It is “how well does it actually work, and what risks does it bring along?” This chapter answers both sides honestly, both what it can do and the points where it still falls short or can easily cross into breaking the law.

One thing you need to understand first is that this is one of the fastest-moving technology areas right now. The tool names and quality levels mentioned here are drawn from sources covering 2025 to 2026. Model versions change very quickly, but the technical limits and legal risks discussed here remain true.

What AI video can really do today

The overall picture of the market is that there are several major players, each strong in a different area. Sora from OpenAI, Veo from Google, Runway, and Kling are names that come up constantly in the conversation. Some are strong on near-film realism, some are strong at controlling the output so it comes out the way you want, some generate sound along with the video, and both price and quality vary quite a bit. The key point for working professionals is this: do not get attached to version numbers or scores that review blogs cite, because those numbers change every few weeks. Look instead at whether the tool can actually do the kind of work your team needs.

In terms of what genuinely works, the strength of AI video is in short clips, social clips that run a few seconds, short ads, b-roll footage, concept illustrations, and making sample ideas before deciding to shoot for real. This group of tasks saves both time and production budget in a visible way. A clear example is a marketing team that wants to test several ad concepts before committing budget to a shoot; they can use AI to create rough samples of each concept to look at first. Or a content team that needs b-roll to go with a clip without buying stock footage every time.

The heart of getting the role right is to see it as an accelerator for concept work and short clips, not a replacement for a shoot that needs high precision. When you place its role correctly, your team gets the benefit without setting expectations in the wrong place.

Quality and the limits that still remain

Even though quality is improving very fast, the technical limits have not gone away, and these limits are what keep AI video from being suitable for certain kinds of work.

First, the clips you get are usually still short. Generating long, continuous video that tells a complete story is still a hard problem. Second, the continuity of objects and physics can still go wrong. Items falling to the ground, flowing water, a hand holding an object: these are points where the model still fails often, because it generates images from patterns it has learned, without actually understanding real physics. Third, the consistency of a character across shots is still a weak point. The same character may look slightly different from shot to shot, which makes it hard to string them together into a single story. And fourth, text within the image often comes out garbled or unreadable, which is a direct problem if the clip needs text, product names, or numbers that must be correct.

The practical result is that work requiring precision, such as demonstrating product use, showing numerical data, or clips with important text, should not yet be left entirely to AI to create. The safer path is to use it to make a draft or component parts, then have a person check and assemble the real work.

This is the section that matters more than the quality question, because getting this wrong does not just make a clip look bad; it can lead to legal liability and damage to your brand.

The first issue is the right to a person’s face and voice, what is called likeness. There are now clear protective laws in many jurisdictions. Taking a real person’s face, voice, or imitated image to make a video without consent risks breaking the law. In the United States there is federal law such as the TAKE IT DOWN Act, which passed in May 2025, criminalizing non-consensual fake explicit videos and requiring platforms to remove them within 48 hours. At the state level there are examples such as Tennessee’s ELVIS Act, which protects name, image, and voice. In Europe, Denmark has amended its law to give people rights over their own face and voice. The caution here is that the rules differ by jurisdiction, and Thailand’s specific legal status on this still needs to be examined separately. So the safest practice is this: do not take the face, voice, or imitated image of a real person, including celebrities and politicians, to make a video without permission.

The second issue is that deepfakes are already being used for fraud; this is not a theoretical worry. Fake video and audio have been used to impersonate people, lure investment, and defraud businesses. The form most directly dangerous to an organization is faking the voice or face of an executive to order a money transfer, which is a risk that has already happened widely. For a business, this means you need an identity verification step outside the video or audio channel for important financial transactions, such as calling back through a known channel, or a multi-layer approval process that does not rely on an instruction from a single clip.

The third issue is copyright. AI-generated video carries the same copyright protection questions as AI-generated images, and you must read each tool’s commercial usage terms before using it for revenue-generating work. Each tool’s terms differ; some grant full commercial rights, some limit them. Skipping this step may mean you cannot use a clip in real work the way you assumed. Note also that the copyright legal framework referenced here is that of the United States; Thailand’s may be different.

The last issue is provenance and watermarks. Some tools add a watermark or Content Credentials following the C2PA standard, which helps identify that a clip was made by AI, but these watermarks can slip off or be removed. The absence of a mark therefore does not mean a clip is genuine. Do not use the presence or absence of a watermark as the sole measure of credibility.

How to use it safely for business and content work

Once you understand both the strengths and the risks, the approach to real use comes down to a few practical principles.

First, set a clear team policy that making video using the face, voice, or imitated image of a real person without permission is prohibited, covering ordinary individuals, celebrities, and public figures alike, because the risk both legally and to brand reputation is too high to be worth it.

Second, put a deepfake prevention process in place on the finance side. Have an out-of-video identity verification step for approving significant sums, and educate staff that an instruction from a clip or from an executive’s voice may be faked.

Third, divide the work by the right type. Work requiring precision, such as product demonstrations, numerical data, or important text, should not be left entirely to AI; use it as a draft, then have a person check and fix it. As for short clips, concepts, and b-roll, let AI act as an accelerator to the fullest.

Fourth, before using a clip for commercial work, read that tool’s usage terms every time, to be sure you really have the right to use it in the context where it will be applied.

Fifth, turn on the watermark or Content Credentials when the tool offers them, and disclose that the content was created with AI when appropriate. This transparency builds credibility with your audience over the long run.

And last, choose tools based on the actual work and your team’s real budget, not on the label of “number one” in a review, because those rankings change quickly and many times carry a whiff of advertising.

Update box: Right now (June 2026)

The information in this chapter is confirmed as of 18 มิถุนายน 2569. This technology area changes fast. The things that move often, and that this chapter deliberately does not pin numbers to, are the version of each tool, the comparative scores per tool, and the price, including free or paid quotas, all of which can change within a few weeks. Before deciding to buy or set a budget, check the current price and capabilities again from each provider’s official page. As for the legal status of deepfakes and likeness rights specific to Thailand, that remains a topic awaiting further verification, and it will be updated when the information is complete.

Next steps

If your team is going to start trying it for real, the recommendation is to begin with low-risk work first, such as b-roll or internal concept clips that have no real people’s faces and do not rely on precise text. Use this round to learn which tool fits your line of work. After that, draft your team’s AI video usage policy covering likeness rights, deepfake prevention on the finance side, and checking copyright terms before commercial use. Having a clear framework from the start lets your team get the benefit from the tools without opening up unnecessary risk.

📚 References

🗓 Verified: 18 มิถุนายน 2569