AI video is developing faster than forecast
If in the period 2023-2024, the AI market mainly focused on image creation, then from 2025 onwards, AI video became a rapidly accelerating field. In just a short time, technology has shifted from creating short-sighted films with many errors to the ability to produce videos with stable characters, natural movements and more complex contexts.
It is noteworthy that the development speed of AI video is not only reflected in image quality but also in the ability to solve problems that were once considered major barriers.
One of them is the consistency of the character. This was once a problem that took many creators weeks to process. The character can change faces, costumes or identifying features after just one scene. However, with the advent of video creation technology from reference images, this problem is gradually being solved.

Virtual singer becomes a "measure" for AI evolution
In Vietnam, the AI/T virtual singer project (Ai Trinh) by musician Thai Nguyen is one of the cases showing this transformation process. According to musician Thai Nguyen's sharing, the first MVs in 2025 are mainly built with a command system (prompt), making maintaining the image of characters a major challenge.
By 2026, when video technology from reference images develops more strongly, products like "Not You Yesterday" begin to show a higher level of stability. AI/T characters can appear throughout many settings in Ho Chi Minh City while still maintaining relatively consistent recognition.
Sharing with reporters, musician Thai Nguyen said: "When starting to make the first MVs for AI/T in 2025, the biggest challenge was not in composing or editing, but in keeping the virtual characters appearing uniform throughout the entire video.
This is one of the special MVs because it is right at the intersection stage: both using the video creation technique with prompts and starting to apply video from images".
However, after only about a year, the technology picture has changed significantly.
New generation AI video platforms now allow creating live videos from reference images, helping to maintain character consistency significantly better. Problems that once took weeks to process can now be shortened to a few hours or even minutes.
This change also gives AI/T project implementers a different perspective when looking back at early-stage products.
According to musician Thai Nguyen, the too rapid development speed of AI makes many products in the period 2025-2026 important observational milestones, clearly reflecting the changes of technology over the years.
From a broader perspective, media expert Hoang Van said that virtual singers show that AI is gradually changing the structure of the creative industry. As tools become smarter, the advantage will no longer completely lie in the ability to operate technology, but shifts to ideas, storytelling thinking and the ability to create emotional value.