Ask an AI video model to show a letter, a newspaper headline or a phone screen, and you get the same thing every time: shapes that look like writing from across the room, and gibberish up close. AI video models can't spell. For a story channel that's a real problem, because so many stories turn on what a document said: the telegram, the court ruling, the text message, the front page.
Synthvids now handles these scenes differently, so the words can actually be read.
Why AI video can't write text
Video models learn what text looks like, not what it says. They're very good at the shape of a page, the glow of a screen, the layout of a newspaper, and very bad at the letters themselves. The longer the scene runs, the more the letters shift and melt from frame to frame.
How Synthvids shows documents you can read
When a scene is about what a document or a screen says, Synthvids makes it a document scene instead of an AI video scene. The document is drawn as a still picture with the exact words on it, readable, in your video's look, and the camera moves slowly across it, the way a documentary shows a letter or a front page.
Synthvids finds these scenes itself when it plans the video: up to 3 in a Short and 6 in a long video, picking the scenes where the words matter most. The rest stay AI video.
Add an image to any scene
In the AI storyboard and pre-render editor, the Storyboard calls these images:
- Click a scene and press Add image.
- If you like, fill in Describe the image or type its exact words (optional), for example "the 1912 front page announcing the Titanic sinking". Leave it empty and Synthvids works it out from the scene's narration.
- Synthvids first looks for a real image whose rights allow reuse, including historical newspaper pages from Chronicling America and items from the Library of Congress and Wikimedia Commons. If none fits the scene, it draws one with the exact words.
A real image gets a credit line in your video description, like the public domain photos Synthvids uses.
Change it, or change your mind
- Find image scenes looks through the whole story for scenes that suit an image.
- Change it makes the scene's image again, from the scene or from what you type.
- Remove image or Use AI video turns the scene back into an AI video scene.
What do image scenes cost?
An image scene costs the same as 7 seconds of AI video: 100 SynthCredits, whether it's drawn or a real archive image. The quote before your render shows it.
FAQ
Why is the text in my AI video garbled?
Because AI video models draw the shape of writing, not real letters. Synthvids gets around it by making text-heavy scenes as still pictures with the exact words, instead of AI video.
Can I show a real newspaper headline in my video?
Yes. Press Add image on a scene, describe the page, and Synthvids searches historical newspaper archives such as Chronicling America for one whose rights allow reuse.
Can I choose what the document says?
Yes. Type the exact words in the box and press Add image (or Change it).
Does every scene with text become a document scene?
No. Only the scenes where what the document says matters, up to 3 in a Short and 6 in a long video. You can add more yourself in the Storyboard.
Does it work for phone screens too?
Yes. A scene about what a screen or a message says can be a document scene too.
Try it
See the plans, then open a story in Storyboard and press Find image scenes.
