Published October 10, 2026
What an online video tool receives
Most AI video services do their work on the provider’s servers. To make your video, they receive everything that goes into it:
- Pictures and screenshots, including products that are not announced yet, the inside of your shop, and the screens of internal software.
- Scripts and documents, such as a staff procedure, a pricing change, or a training manual you would rather keep in-house.
- People, when photos show customers or staff.
- Your voice, when a tool offers to clone it from a recording.
For a public ad, much of that ends up public anyway once the video is posted. For training videos, internal announcements, and anything still in development, the drafts and source files are often the sensitive part.
Questions to ask an online video tool
If you use a cloud service, its terms of service and privacy policy should answer these. If they do not, ask.
- How long are uploaded files, scripts, and finished videos kept, and can you delete them?
- Are uploads used to train or improve the provider’s models, and can you switch that off?
- Who owns the finished video, and what may you do with it?
- If you clone a voice, who can use that voice model, and what happens to the recording?
- Do credits, minute limits, or watermarks apply to the plan you can afford?
- Do you have permission from everyone who appears in the pictures?
How local video AI works
A local video tool keeps every step on your own computer. A language model plans the scenes and writes the narration, a speech model reads it aloud, and a video engine renders the file, all on your processor and graphics card. Your pictures, documents, and drafts stay in a folder you control, and nothing needs an internet connection once it is set up.
The Video Studio Edition works this way. It runs from an Evolve AI drive on your Windows computer: the AI that writes, the voices that speak, and the engine that renders all run locally, with no account and nothing uploaded.
What local video AI can do today
Using Video Studio as a concrete example:
- Plan a video from a sentence or a document. Describe what you need, or open a Word, PDF, or text file, and the local AI writes every scene and its narration. On our development computer, planning a 30-second video took about 10 to 15 seconds.
- Narrate it with natural voices. Twenty-eight American and British English voices that you can hear before you choose, with a pronunciation list for names the voices get wrong.
- Add captions and music. Captions follow the voice, with .srt and .vtt files for YouTube and Facebook, and the app composes original music for each video.
- Fit every shape. Wide 16:9, tall 9:16 for Reels and Shorts, and square 1:1.
- Point at what matters. Highlight boxes and camera moves on screenshots, for how-to and help videos.
- Make audio only. On-hold messages and voiceovers as MP3 files.
- Render without limits on count. No credits or per-video fees. On the development computer, videos of 19 to 32 seconds rendered in 53 to 68 seconds.
What it cannot do yet
Local video AI has real limits, and they are worth knowing before you buy:
- No talking avatars or live-action generation. Videos are built from your pictures, screenshots, AI-created pictures, and text scenes.
- No editing of existing video clips in Video Studio version 0.1.
- No voice cloning. The voices are synthetic and are meant for your own content.
- English narration only for now.
- AI pictures need the right hardware: an NVIDIA graphics card and at least 12 GB of memory. Without one, you can still use your own photos and screenshots.
- Speed depends on your computer. Videos render on the processor, so a slower computer takes longer. Video Studio has been tested so far on one Windows 11 laptop.
- You post the finished files yourself. Upload each video to the sites you choose.
The AI drafts from what you give it and can get details wrong, so read every scene before you publish. Video Studio flags prices, dates, and offers your brief did not mention.
When an online tool fits better
Online video services earn their place for some jobs. Choose one when you need a talking presenter, edits to footage you filmed, narration in many languages, or several people working on the same project from different computers. They also run on a Mac or a Chromebook, and they need no particular hardware.
A sensible split for many businesses: use an online tool for public material where nothing sensitive goes in, and a local tool for training, internal announcements, unreleased products, and anything with customer or staff details.
A private workflow, step by step
- Gather your material in one place. Photos, screenshots, your logo, and any document the video should follow.
- Describe the video. One or two sentences on the audience, the message, and the call to action.
- Review the plan. Read and listen to every scene, fix names and numbers, and adjust the order.
- Render and check. Watch the whole video with the sound on, and once more with it off to check the captions.
- Publish only the final file. Your drafts, source pictures, and scripts stay on the drive.
Using AI voices honestly
Synthetic voices are now natural enough that a listener may assume a real person is speaking. If a viewer could mistake the voice for a real person, say that the video uses an AI voice. Use AI voices only for your own content, never to imitate a real person, and use only pictures and music you have the right to use.
Want to hear the voices and watch unedited videos made on the development computer? Visit the Video Studio showcase, then see the drive prices.