Shipping
Vault X
Desktop Utilities
Encrypts and hides your files, photos and notes behind a PIN or Windows Hello, completely offline.
- Platform
- Win 10 / 11
- Cipher
- AES-256-GCM
- Price
- Free
Products/AI Applications
Release pendingEngine complete. Installer pending.
Three hours of you. Eleven clips worth posting.
Watches your long video the way an editor does and cuts it into vertical shorts with burned-in subtitles, entirely on your own GPU.
Why it exists
OverviewThe problem
Every cloud clipper starts the same way: upload four gigabytes, wait behind strangers in a queue, pay monthly, take the watermark, and hope the terms of service still say what they said last year. Meanwhile your footage sits at rest on someone else's storage.
What Veldora does
Veldora hears every word, knows who is speaking, scores every second for hooks and payoffs, and exports the survivors as vertical shorts. It runs on your own GPU and its engine is bound to localhost, a socket with no route to the internet. Unplug the ethernet cable mid-render and nothing happens. That's the feature.
What it does
6 capabilitiesFinds the moments that land
Scores every second of the transcript, hooks, punchlines, story beats, and keeps only what survives the threshold.
Hears who is speaking
Whisper large-v3 transcribes, pyannote separates the voices, NLLB-200 translates. Hindi in, English subtitles out.
One warm GPU, ten steps
A persistent Python engine holds every model in VRAM. Load the models once, not once per step.
Reframes for every format
9:16 shorts, 1:1 posts and 16:9 clips, with word-level subtitle timing burned in or left off.
Reads your machine first
On first launch it checks your GPU, VRAM, RAM, cores and free disk, then recommends model sizes that will actually fit.
Nothing leaves
No accounts, no telemetry, no watermarks, no monthly invoices, no uploads. Total data sent: zero bytes.
Datasheet
Specifications & requirements- Platform
- Windows 10 and 11
- Requires
- NVIDIA GPU with CUDA. 8 GB VRAM comfortable.
- Takes in
- A YouTube URL, or MP4 / MKV / MOV / AVI / WEBM
- Gives back
- MP4: 9:16 shorts, 1:1 posts, 16:9 clips
- Speech
- Whisper large-v3 · pyannote · NLLB-200
- Built with
- WinUI 3 · .NET · Python · FFmpeg
- Network
- Engine bound to 127.0.0.1. Zero bytes leave the machine.
- Status
- App and engine complete. Installer not built yet.
- Price
- Free when it ships
Questions
FAQCan I download it today?
No. The app and the AI engine are finished and working, but the installer is not built. Packaging a bundled Python runtime and a multi-gigabyte model set into a single Windows install is its own project, and it is the one I am on.
Do I need an NVIDIA GPU?
Yes. Veldora runs the models locally with CUDA. 8 GB of VRAM is comfortable; more is faster.
Is my footage uploaded anywhere?
No. The AI engine listens on 127.0.0.1 and has no route to the internet. The only thing ever downloaded is the AI models, once.
Release pending
Veldora isn't out yet
Not released yet. The app and the AI engine are finished; the installer is the one thing left, and it is the hard part: a bundled Python runtime plus multi-gigabyte models as a single Windows install.
Also on the bench
Related productsShipping
JerseyFactory-DTF
Creative Tools
Turns a list of player names and numbers into a print-ready gang sheet. Built for the press floor, not the design studio.
- Platform
- Windows 11
- Film
- 22" / 24" / 30"
- 100 players
- 2 min