Skip to content
Nguyen Long Vu

Studio

Explainer videos, checked frame by frame and handed over editable

Scripts, voice, scenes written in code, automated checks on every render, and a CapCut project you can still change yourself. English or Vietnamese.

AI use is labelled per reel

Pin Yếu, the character from the teaser, walks across the page. His walk cycle is the same code as in the video.

Pin Yếu, a horror teaser in Vietnamese

My own pitch for a short-form horror channel: a phone battery drains while a boy walks the fourth floor of an old apartment block looking for the fuse box. The blackout at the end is the story, not a render fault.

Pin Yếu demo: a silhouetted character walks a dark corridor with a flashlight while a phone battery meter drains in the corner.

Reel log

ReelFormatLengthScriptVoiceVisualsSound
Pin Yếu teaser16:9 · 1080p0:55Mine, VietnameseSynthetic voiceDrawn in code, no generated imagesSynthesized in code, no samples
Client explainer16:9 · 1080p4:30Client'sSynthetic voice2D rig in code, some AI-generated shotsNarration

The video is not public. With the client’s consent, these are its QA facts: A 270-second explainer rendered 6,482 frames with 0 black frames. After a voice swap, tests caught a 1,313-frame drift in the narration and I re-aligned it to under 1 frame.

Client deliveries

Education and storytelling channels · 10/2026

A 4:30 explainer, re-voiced and handed over as an editable project

4:30
running time
178
Subtitle lines, 27 keywords highlighted
16
Scenes in the editable CapCut project
6,482
Frames checked, 0 black

The client changed the narration to a new voice. I re-timed the existing scenes, real-footage and chibi, to the new track and set the music to duck under the voice.

Captions came as an SRT in the client’s house style: Montserrat, with keywords in bold cyan, plus a clean version without highlights.

The CapCut project was rebuilt scene by scene so the client can change one part without touching the rest. When the international CapCut build refused to open the project because of a file path, I fixed the paths and re-sent it. Two scenes went back once for a shirt colour that did not match, and were re-delivered.

At the end I walked the client’s team through running the same pipeline on a call, so the next videos do not wait for me.

The client is not named and the video is not published.

Van Data Team · 08/2026

A news-video pipeline that only says what the source supports

  • Narration is written from the source article and limited to what it states; claims without support are dropped before voice.
  • A word-alignment gate stops a render whose captions drift from the voice.

Built on a third-party video framework; the parts listed are mine.

Model sheet

Pin Yếu model sheet: the main character from front, side and back, a six-phase walk cycle, and poses for waving, fear, holding a lamp and blinking.
The Pin Yếu character: front, side and back views, a six-phase walk cycle, and the poses the episode uses. Every frame of the rig is code.

How a video is made

Step 1 of 6

A short call and a written brief, then a script the client approves before any voice is recorded.

  1. Normal run01DoneA short call and a written brief, then a script the client approves before any voice is recorded.
  2. 02DoneThe voice track is generated or recorded, and speech recognition gives each word a timestamp. Captions and cuts follow those timestamps.
  3. 03DoneScenes are written in code with Remotion: character rigs, camera moves and text, so a change to the script re-renders instead of re-editing.
  4. 04DoneEach render is checked for duration, frame count, black frames and loudness. The client gets the MP4, the captions and a CapCut project they can still edit.
  5. Something breaks05FailedOn one delivery the narration was replaced with a new voice. After the swap, the timing test failed: the narration had drifted 1,313 frames from the scenes.
  6. Recovery06DoneI re-aligned the narration to under 1 frame, re-rendered, and the checks passed before anything reached the client.

What I built, and what I build on

Mine

Custom scenes and character rigs, word-timed captions, the CapCut exporter and the automated checks (duration, frame count, black frames, loudness) that run on every render.

Built on

Remotion for rendering, and a third-party video framework and a licensed template package as the starting point of the pipeline. Their code is not redistributed.

What you get

  • MP4 in 16:9, 9:16 or both
  • Subtitles as an SRT file
  • A CapCut project you can edit
  • A QA sheet: duration, frames, black-frame scan, loudness

Writing and localization

The same checks apply to text: I edited the Vietnamese editions of vandatateam.com and vanaxity.com against the English originals, and the copy on this site passes a lint for machine-sounding phrases.

See the record →

Tell me what you need built

A few lines are enough: what runs today, what should change, and by when. For a role, paste the job link.
What do you need?
When do you need it?

Your message goes to my inbox and my lead sheet. No newsletter, no third-party tracking.