Generative AI
We produce content with generative AI: promotional videos, educational training, realistic UGC and lipsync for AI characters.
REQUEST A DIAGNOSTICWhat generative AI is
Generative AI creates new content, such as images, video, voice or text, from a description or reference material. Applied to production, it makes it possible to produce pieces that used to require a shoot, a voice-over or manual editing, and to get variants of the same piece without repeating all the work.
Ways to generate video
Text to video
The video is generated from a description. It gives more freedom and less control over the result.
Image to video
A starting image is animated, which fixes the composition, the light and the features. The better that image, the fewer attempts are needed.
Talking avatar
A synthetic presenter who reads a script. It can be built from a real person, with their consent, or be a generated character.
Lipsync
The mouth area is regenerated to match new audio. It is used to dub a video or change its voice-over.
What we produce
Promotional videos
Pieces to present a product, a service or a brand.
Educational training
Content to explain processes, products or subjects.
Realistic UGC
Videos that look like content recorded by a user.
Lipsync for AI characters
Generated characters who speak, with lips and voice in sync.
Examples of what we have already done
Promotional videos
A series of short videos to present products or services, with variants for each channel.
Training
Video lessons with a synthetic presenter for courses or employee onboarding.
Social media content
Pieces in a user-generated content format for social campaigns, stating that they are AI-generated where appropriate.
Versions in other languages
A video with synchronized voice and lips (lipsync) in several languages, without recording it again.
Product images
Images of a product in different settings, without a photo shoot for each one.
Internal communication
Regular announcements to staff with a synthetic presenter.
When it helps
When content has to be produced continuously, when the same piece must be adapted to several formats or languages, or when you want to test ideas before investing in a traditional production.
Where a person is still needed
- Generated content needs review before it is published: a model can produce errors of detail or continuity.
- Using the image or voice of a real person for commercial purposes requires their express consent.
- Creative direction remains human: the model executes, it does not decide what has to be told.
- An avatar is convincing in short shots of direct speech. In long shots, repeated gestures and blinks show, so it is edited with cuts and supporting shots.
- Hands, teeth and interaction with objects are still weak points. Showing a real product in use may require a shoot.
- Lipsync works best with clean audio and a single person facing the camera.
- The EU AI Act requires content to be labeled as artificial when it imitates real people, places or events in a way that would seem authentic.
FREQUENTLY ASKED QUESTIONS
What is generative AI?
It is the kind that creates new content, such as images, video, voice or text, from a description or reference material.
What content does Devontic produce with generative AI?
Promotional videos, educational training, realistic UGC and lipsync for AI characters.
Can generated content be published without reviewing it?
It is not advisable. A model can produce errors of detail or continuity, and using the image or voice of a real person requires their consent.
Who reviews a piece before it is delivered?
We review it and the client also reviews it before publishing.
Can you tell that a video was made with AI?
It depends on the shot. In short shots of direct speech it can go unnoticed. In long shots, with moving hands or with objects, it is easier to notice. That is why the script and the edit are adapted to what the technique does well.
What does AI-generated video still not do well?
Showing a real product in use, keeping the same character across many scenes without reference material, and sustaining long shots without cuts.
Can I use my image or that of someone on my team in an avatar?
Yes, with that person's express consent. In Spain, using someone's image or voice for advertising or commercial purposes without their authorization is an unlawful interference with their right to their own image.
Does a video have to be labeled as AI-generated?
When the content imitates real people, places or events in a way that would seem authentic, yes: the EU AI Act requires whoever uses it to disclose that it is artificial. For clearly creative or fictional works, the obligation is limited to disclosing it in a way that does not spoil the work.
What does the cost of a generative AI video depend on?
On the total length, the number of variants and languages, whether a custom avatar is needed, and how many attempts each shot takes. In generated video, a good part of the cost is repeating until a valid shot is achieved.
From the idea to the piece
We start with a proof of concept on real data and only scale if the value is demonstrated.