
Making a 5-Minute AI Short Film Alone With Claude and MiniMax H3: Video Costs of Only About 100 Thousand VND
Vũ Khôi
Tác giả tại Wotech Electronics Blog

Making a 5-Minute AI Short Film Alone With Claude and MiniMax H3: Video Costs of Only About 100 Thousand VND
A Chinese creator shared how to make a 5-minute AI short film, with characters who open their mouths and speak, for video costs of only about 100 thousand VND — according to a post from the Hỏi Dân IT Official page retelling the creator's article on Xiaohongshu (account Miller, @TuyetloanPB1840). At the official price of Seedance 2.5 for the same length, it would cost over 5 million VND, a difference of nearly 50 times, per the post's comparison. The resulting film, according to the post, looks quite polished: characters speaking, expressing, and moving smoothly, with beautiful shots — made by the creator alone, with no expensive graphics card. All cost figures and interpretations in this article belong to the original article's author and the retelling post, and have not been independently verified by Wotech.
Quick answer: According to the original article, one person, without buying a graphics card, can make a short film whose characters speak: each video segment costs about 0.2 yuan, a 5-minute film under 30 yuan; finer post-processing runs about 0.3 to 1.5 yuan per segment. The post converts the film's video-generation cost to "only 100 cành" (about 100 thousand VND), versus the official Seedance 2.5 price from Volcano for the same volume of about 1,360 yuan (over 5 million VND) — nearly 50 times apart. The division of labor: screenwriting and scene breakdown with Claude (DeepSeek works too); art, characters, scenes, and props with GPT Image in ChatGPT (or Jimeng, Nano Banana); turning images into moving video with dialogue and sound with MiniMax H3 running on RunningHub; editing with CapCut (剪映 in Chinese). The creator plays producer: setting the story, picking shots, making the calls. The order is one sentence: fix the cast first, then shoot, then edit into a film. The author also shared 8 prompt sets, the complete H3 workflow, and 3 Claude Skills, from tips for keeping characters' faces consistent to fixing broken scenes. The original article, according to the post, reached nearly 3 million views in under a week.
MiniMax H3 is MiniMax's video generation model, run here on the RunningHub platform to turn still images into moving video with dialogue and sound. Seedance is ByteDance's (Volcano Engine) video generation model family; the post uses the official Seedance 2.5 price as the cost-comparison benchmark.
Four "weapons" in the workflow
| Role | Tool per the article | Job |
|---|---|---|
| Screenwriter, scene breakdown | Claude (DeepSeek works too) | Writing the script, breaking down each scene |
| Art | GPT Image in ChatGPT (or Jimeng, Nano Banana) | Designing characters, settings, props |
| Image-to-video | MiniMax H3 on RunningHub | Turning images into moving video with dialogue and vivid sound |
| Editing | CapCut (剪映) | Cutting smoothly into a complete film |
| Producer | The creator themself | Setting the story, picking shots, making the calls |
One person is a whole crew, as the article puts it: the order is one sentence — fix the cast first, then shoot, then edit into a film. Both animation-style dramas and realistic live-action-style short dramas can use this workflow; the visual style is chosen right in the prompt.
How is it nearly 50 times cheaper?
The cost comparison in the post and the original article:
| Item | Figure per the cited source |
|---|---|
| Each video segment (Claude + H3 workflow) | About 0.2 yuan |
| A 5-minute film with speaking characters | Under 30 yuan; the post converts the video cost to "only 100 cành" |
| Extra fine post-processing per segment | About 0.3 to 1.5 yuan |
| Same volume with Seedance 2.5 (official Volcano price) | About 1,360 yuan (over 5 million VND) |
| Difference | Nearly 50 times, per the post |
The point of the comparison, according to the post, is not only cheapness: the film still has characters speaking, expressing, and moving smoothly, with beautiful shots — while the creator worked alone with no expensive graphics card.
What did the author share?
According to the post, the author went all out: 8 prompt sets, the complete H3 workflow, and 3 Claude Skills — from tips for keeping characters' faces from changing scene to scene (character consistency) to "firefighting" broken scenes. From script to editing, every step has searchable, copyable prompts, 8 segments in total; the tuned H3 workflow is published alongside, ready to open and use. No wonder the article, according to the post, reached nearly 3 million views in under a week.
Frequently asked questions
1. How long is the AI short film, and what did it cost? According to the original article, a 5-minute film with speaking characters: each video segment about 0.2 yuan, the whole film under 30 yuan; the post converts the video cost to about 100 thousand VND. The figures have not been verified by Wotech.
2. Which tools does the workflow use? Per the article: Claude for script and scene breakdown (or DeepSeek), GPT Image in ChatGPT for art and characters (or Jimeng, Nano Banana), MiniMax H3 on RunningHub for image-to-video with dialogue and sound, CapCut for editing; the creator plays producer.
3. Why is it nearly 50 times cheaper than Seedance 2.5? Per the post's comparison, the same video volume at the official Seedance 2.5 price from Volcano is about 1,360 yuan (over 5 million VND), while the Claude + H3 workflow costs under 30 yuan for a 5-minute film.
4. Which film styles can this workflow produce? According to the article, both animation-style dramas and realistic live-action-style short dramas work; the visual style is chosen right in the prompt.
5. Did the author share materials? According to the post, yes: 8 prompt sets, the complete H3 workflow, 3 Claude Skills, character-consistency tips, and broken-scene fixes — the tuned H3 workflow is published alongside, ready to open and use.
Source: post by the Hỏi Dân IT Official page (18/11/2026) retelling the article by Miller (@TuyetloanPB1840) on Xiaohongshu. The cost figures, workflow, and interpretations in this article belong to the original article's author and the retelling post, and have not been independently verified by Wotech.
Bài viết liên quan

18 tháng 11, 2026 · 6 phut doc
FPT Software Chairwoman: US Clients No Longer Want to Pay by the Hour, Company Shifts to Outcome-Based Pricing
Artificial intelligence is forcing FPT Software to transform its traditional software outsourcing model: large clients in the US now apply 100% outcome-based payment, and the company expects the Time…

18 tháng 11, 2026 · 6 phut doc
Meta Expands Muse to Australia: The 6th Country, Real-Time Voice, and Local App Integrations
On 17/11/2026, Meta expanded Muse to Australia — the 6th officially supported country, just over 2 months after its US launch. According to a post by Hồ Hoàng Tiến (18/11/2026), the notable point is …

18 tháng 11, 2026 · 7 phut doc
AI Agent Costs Grow With the Square of Steps: 4 Layers to Cut Token Spend Before Switching to a Cheaper Model
You don't pay for answers — you pay for how many times you re-ask the same thing. That is the central argument of Hoang Viet AI in a post in the comailo group (18/11/2026): an agent running 40 steps,…