When I thought about sharing AI news through short videos, the first problem I faced wasn’t editing, but ‘who to deliver it to and what stories to tell.’ There are various announcements every day. However, even if I introduce model names and technical terms as they are, I myself end up thinking, ‘That sounds amazing, but what does it actually mean?’
That’s why I created a fictional guide named ‘Sora Aose.’ I set her up as someone who translates difficult topics into plain language and asks questions about things that are concerning. Her appearance features short black hair with blue highlights. I’ve also created her voice and video templates. However, the most useful thing I found after creating her wasn’t her face or voice, but the criteria of ‘which news would Sora choose?’
This article is not just about having an AI create a character image and calling it a day. It is a record of the criteria I set to keep going every day and the places where I, as a human, perform checks. As of October 3, 2026, the production mechanism is in place, but whether the posts will continue steadily and whether readers will follow is something I will verify from here on out.
At first, I only thought about ‘making AI news every day’
It started with a short video that suddenly appeared on YouTube
Collect news, turn it into a short script, and make it a vertical video. Looking at the flow alone, it seems simple. However, when I looked at the candidates, the distance between the topics was scattered, such as funding, research papers, updates for developers, and features to use on smartphones. If I choose one every day, I can’t decide just by picking ‘the biggest announcement of the day.’
I gave Sora a focus: ‘someone who cares about stories that might change our lives, work, and smartphones starting tomorrow.’ I prioritize changes in familiar products and stories that make you want to ask, ‘What can I do if I use that?’ I pass on funding news that only mentions amounts and detailed technical stories that aren’t yet linked to the actions of ordinary people.
By setting this up, I can ask the AI not only to ‘collect 8 topics’ but also to provide ‘Sora’s first reaction’ for each candidate. For example, whether she was surprised or curious about the conditions for use. Looking at that one line, I make the final choice of the subject. The character became a yardstick for choosing news.
Before her appearance, I decided how she would speak when she didn’t understand something
Someone somewhere once said, ‘By setting a character for AI, you develop an attachment and use it more.’
Sora is set to be 24 years old and from Sapporo. She is curious and paraphrases difficult words. However, I didn’t make her a teacher who knows everything. She is someone who doesn’t pretend to know things that cannot be verified by the source.
For AI news, the conditions of ‘what it can now do’ are important. Is it limited to certain users, does it require a fee, or is it still in the announcement stage? If I talk about it as if anyone can use it from today while that part remains ambiguous, it spreads misunderstandings even in a short video. Thinking about Sora-like phrasing and verifying the facts were a set.
I made her voice and expressions capable of showing a little surprise, doubt, and understanding. The blue sphere in the video moves in sync with her voice. However, even with a voice, it doesn’t do the fact-checking for me. I am responsible for the final selection of the topic and the verification of numbers and terms of service.
To make it every day, I don’t think about the video from scratch
I decided on a template of ‘sphere and audio’ for the visuals. I decide everything in advance, from the length of the subtitles and the number of characters in the script to the daily accent color, music, and how she introduces herself at the end. This is because if I think about a completely new production every time, the time to verify the content of the news decreases.
My workflow is to look at the candidates in the morning, choose one, check the completed script and video, and have it corrected if necessary. I can also create the Threads text that follows the video separately, but I don’t use the fact that I watched the video as approval for the text. I read the text as text.
I stumbled here once. Even if there is a production mechanism, if I can’t see ‘what to choose from’ on the initial news selection screen, I can’t make a judgment. I revised it to show not just the candidate number, but also the headline, what changes, and Sora’s first reaction together. Even if the AI collects candidates, it stops if it’s not displayed in a way that a human can choose.
What has actually been done and what will be verified from here on
I have created Sora’s settings, the criteria for news candidates, the video template, the audio settings, and the verification flow. On October 3, I also exported a video of about 34 seconds. On the other hand, as of the same day, that video was waiting for identity verification. I have also passed tests for the automatic posting flow, but whether it will run without problems every day on a real account is a separate verification.
I cannot yet say that ‘creating a character will boost your reach.’ What I can say now is that deciding on the character’s interests and tone made it easier to select news and write scripts.
For those thinking the same thing
If you are going to create a guide without showing your face, I think it is easier to proceed if you write down just three things before generating images.
ã»What kind of news does this person feel is ‘relevant to them?’
ã»What would they ask about if they didn’t understand something?
ã»What kind of stories would they pass on?
Once you have done that, try asking the AI, ‘Which story would this person choose? Give me the reason and their opening line.’ That was when I first felt that Sora had become a guide that helps with daily decisions, rather than just being a decoration.
ããã£ããã¤ã³ã¹ã¿è¦ãã¦ä¸ããã¾ããã¾ã ã¾ã å®é¨ä¸ã
Originally Appeared Here
