Key takeaways
- Click AI rewrite on a saved item and Beav opens a new chat with a reference to that item and the rewriting skill selected. Add what to keep, what to change and whether you want text, images or video, then send it yourself.
- The AI first describes how the original is actually put together, then proposes how the new piece maps onto it, and waits for you to approve the plan. When the language, a character or the product changes, labels, dialogue, imagery and sound are considered together.
- Targeted edits such as "change only page two" or "replace the voice" reuse the accepted assets that are not affected. Data and demonstrated results from the original are never facts about your product.
The usual way to adapt a piece of content is to paste it into an AI and ask for "a different version". What comes back is either a synonym pass that still reads like the original, or something rewritten so heavily that whatever made the original work is gone.
What makes a piece work is rarely a few words. It is the relationships: which object a label on the image points at, whether two things are compared side by side on the same page, how the action in one comic panel carries into the next, which line of the voiceover a visual change lands on. A good rewrite keeps those relationships and changes the language, characters, product and facts. Beav's AI rewrite is built around that idea.
Who it is for
- Creators with a reference that performed well who want a version for their own product or brand.
- Anyone turning the same image post or short video into another language, say from French to Spanish.
- Teams that, once a version is done, keep needing "just change this page" or "just swap the voice".
How to use it
Save the reference to your Beav knowledge base first. Rewriting runs on the chat model and uses model quota according to the usage rules.
- Click AI rewrite on the item. Open the reference in the knowledge base and click AI rewrite at the top right. Beav opens a new conversation with a prefilled prompt, a reference to the current item and the rewriting skill selected. Clicking the button does not send anything.
- Say what to keep, what to change and what to deliver. For example: "Keep the four-panel single page, characters and reading order; only change French to Spanish." Then say whether you want text, images or video. A video source does not mean you want a voiceover script, and the AI does not pick the deliverable from a "video" or "image post" label alone.
- Review, send, and read the plan first. Based on the text and media it actually read, the AI explains how the original is organized and proposes how the new piece will correspond. By default it waits for your approval. If you already told it to go ahead, or you are continuing an approved task, that instruction still holds.
- Let it produce. The confirmed format carries into image and video production, so you don't have to restate it at every step.
- Scope your edits. Ask for "change only page two" or "replace the voice but keep the visual relationships". The AI should reuse the accepted assets that are not affected and check the revised result. After a language or voice change, recheck the timing of subtitles and visual changes.
What to keep and what to change
The rewriting guidance looks at four kinds of relationship: how text maps to objects, whether a comparison can be read at a glance, how actions connect between comic panels, and which words a visual change responds to. These are usually what make the original effective, so they are worth keeping.
What changes is the language, the characters, the product and the facts, and they change as a set. A new product affects the labels on the image, the characters' dialogue, what is in the frame and the voiceover. Rewriting only the body text leaves the words and pictures out of step.
How it compares with pasting the text into ChatGPT
| Paste the text into ChatGPT | Beav AI rewrite | |
|---|---|---|
| Starting point | Text you paste or screenshots you upload | An item already saved in your knowledge base, attached to the chat as a reference |
| First move | Usually a rewritten version straight away; ask if you want an analysis first | Describes how the original is organized, proposes a mapping, waits for approval |
| What changes | Mostly the text | Labels, dialogue, imagery and sound together, delivered as text, images or video as you asked |
| Targeted edits | You can ask for changes; keeping the rest intact is on you to watch | Name the part to change; unaffected accepted assets are reused |
| Strengths | Always at hand, no setup, fast for polishing a paragraph | The reference is already in your library and you want a finished image post or video |
| Cost | Your ChatGPT plan | Beav's chat model, using model quota according to the usage rules |
If you just want a paragraph reworded, pasting it into ChatGPT is the quickest route. Beav suits a different case: the reference has images, visuals and pacing, and you want to keep those relationships while switching to your own language and product.
Limits and notes
- The original's data is not your evidence. Figures, test results and demonstrated effects from the reference cannot be presented as facts about the new product. If you lack evidence, add your own material first.
- The AI can only vouch for what it saw. If pages or media cannot be read, it should say what it actually observed. A title, OCR or a written plan does not show it understood visuals it never saw, and does not replace an agreed image or video deliverable.
- AI rewrite only prepares the chat. Look over the prompt before sending, and check the results yourself.
- A rewrite learns structure and expression. Get the author's permission before using the original's images, video clips or voice.
Next steps
- Break the reference down before rewriting: turn a video into a shot-by-shot script.
- Writing technique such as structure cards and removing the AI tone: viral post breakdown and rewriting.
- Full reference: keep the relationships that make the reference work.
Applies to Beav desktop 2.9. Checked 2026-10-22.
