Free Intensive Nov 21–22Free 2-day hands-on Vibecoding Intensive on Nov 21–22 at 16:00 MSK (Sat–Sun) · Live Cohort 3 starts Nov 23Join Free →Cohort 3 (Nov 23)

An Anime-Style Minute-Long AI Clip Built in 8 Hours for $55 and 20 Prompts

The creator, going by yarila2607, made a fan-made clip of Krovostok's "Angry Pigeons" that runs over a minute in less than 8 hours. Claude handled most of the storyboarding, GPT Image and Nano Banana generated the characters, and Vivix and T1 from TopView produced the video. The total cost was about $55, and the creator needed roughly 20 prompts. The takeaway: a single person can now produce a stylistically consistent video longer than a minute in one evening.

AI-processed from TG: @cgevent (Метаверсище и ИИще); edited by Hamidun News
An Anime-Style Minute-Long AI Clip Built in 8 Hours for $55 and 20 Prompts
Source: TG: @cgevent (Метаверсище и ИИще). Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceTG: @cgevent (Метаверсище и ИИще) ↗
Editor-in-Chief
Key metric #1Claude and the images cost about $40, and the video footage about $15
Key metric #2Vivix lacked dynamism, though it sometimes delivered decent results
Key metric #3Claude came up with most of the shots itself, though some required corrections
An Anime-Style Minute-Long AI Clip Built in 8 Hours for $55 and 20 Prompts
Featured Video / Demo
Click to load the embedded player (privacy-enhanced youtube-nocookie mode) or read the full breakdown below.Watch on YouTube ↗

Long AI videos have remained the weak spot of generative video: ten-second clips look impressive, but at the one-minute mark characters drift, the style shifts, and everything falls apart into disconnected fragments. That is why a fan-made clip for the song "Angry Pigeons" by Krovostok, covered by the channel "Metaverseshche i IIshche" in its "Neuroprozharka" column, is interesting not for the music but for its production numbers: more than a minute of coherent video, less than 8 hours of work, and a budget of about $55.

What the clip is made from

The author, who signs as @yarila2607, names his toolset. Claude handled most of the storyboarding. Characters were generated in GPT Image and Nano Banana. The video was assembled in Vivix and in the T1 model from TopView. Fable 5.2 is also mentioned in the list, but the post gives no details about its role, so we won't speculate.

The division of labor looks like this. The idea and direction stayed with the human. He picked up the style from short-form content and doesn't remember the original creator. The language model turned the concept into a sequence of shots: what appears in each frame, who does what, and how the shots connect. Then image models drew the heroes and scenes, and video models brought them to life. According to the author, Claude came up with most of the shots well on its own, but in places it needed help.

What it cost

The author gives specific figures. Claude and the images from GPT and Gemini cost about $40. Generating the video footage added around $15. The total came to $55. On the human side, it took about 20 prompts and under 8 hours.

Comparing this to a conventional clip would be misleading: an amateur project has no crew, no actors, and no licensing for the track. But the order of magnitude is telling. Previously, a video with recognizable characters and a storyline required a team and weeks of work. Now one person fits it into a working day and the cost of a few lunches.

What is honestly limited here

First, the author himself admits that Vivix falls short on dynamism. The video model produced very decent results in some scenes but not in all of them, which is a typical picture: out of ten generations, two or three are usable. How many takes he discarded isn't clear from the post, and 20 prompts do not equal 20 generations.

Second, this is not an independent test or a benchmark. It is the story of one enthusiast, and quality can only be judged from the video on YouTube, which is linked in the post. The cost figures are also the author's own claims.

Third, the main goal of the project was not the clip but testing a platform the author has been building for about a year. He needed to put it through its paces on long videos. By his account, the latest model improvements finally gave him acceptable quality on video longer than a minute. What exactly the platform does is not stated in the post.

What to take from this

For those building products around generative video, two observations are useful. First, the chain of a language model producing the storyboard, an image model fixing the characters, and a video model animating them can already be assembled into a working pipeline. Anchoring characters with separate images helps keep their appearance stable from shot to shot. Second, the bottleneck has shifted. What has become expensive is not rendering but direction: the idea, taste, and shot selection remain with the human.

For producers and studios, the conclusion is simple. Clips, commercials, and teasers longer than a minute are no longer the privilege of those with a shooting budget. Competition is moving toward those who can frame a task well for models and who can pick out the successful shots. For now, video models still trail live-action footage in dynamism, but judging by this example, the gap is closing faster than the cost is falling.

Story

Key Questions Answered

What the clip is made from?

The author, who signs as @yarila2607, names his toolset. Claude handled most of the storyboarding. Characters were generated in GPT Image and Nano Banana. The video was assembled in Vivix and in the T1 model from TopView. Fable 5.2 is also mentioned in the list, but the post gives no details about its role, so we won't speculate. The division of labor looks like this. The idea and direction stayed with the human. He picked up the style from short-form content and doesn't remember the original creator.

What it cost?

The author gives specific figures. Claude and the images from GPT and Gemini cost about $40. Generating the video footage added around $15. The total came to $55. On the human side, it took about 20 prompts and under 8 hours. Comparing this to a conventional clip would be misleading: an amateur project has no crew, no actors, and no licensing for the track. But the order of magnitude is telling. Previously, a video with recognizable characters and a storyline required a team and weeks of work.

What is honestly limited here?

First, the author himself admits that Vivix falls short on dynamism. The video model produced very decent results in some scenes but not in all of them, which is a typical picture: out of ten generations, two or three are usable. How many takes he discarded isn't clear from the post, and 20 prompts do not equal 20 generations. Second, this is not an independent test or a benchmark. It is the story of one enthusiast, and quality can only be judged from the video on YouTube, which is linked in the post.

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Private AI Builders Community · Vibe Club

Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews

Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.

Join Vibe Club →Explore Hamidun Academy
What do you think?
Loading comments…