La mejor IA de VIDEO libre (increíble para YOUTUBE) 🤯 Nuevo MiniMax H3
This video demonstrates MiniMax H3, a free open-source video AI model that generates commercial-quality 2K videos with superior text rendering and audio capabilities at approximately 8-10 cents per second. The creator showcases its versatility through multiple examples including product commercials, scene editing, and motion transfer, positioning it as competitive with or superior to Runway Sideance 2 while being significantly more affordable.
Summary
The video introduces MiniMax H3, a new multimodal AI video generation model that accepts image, video, and audio inputs to produce videos. The creator demonstrates that the entire opening sequence shown was AI-generated, establishing the model's capability to create entirely synthetic or edited content indistinguishable from reality.
The model ranks second in text-to-video generation with audio according to Artificial Analysis rankings, third in image-to-video generation, and leads in editing capabilities. It generates in 2K resolution and costs approximately 8-10 cents per second, which is substantially cheaper than competing models like Sideance 2 (which costs 250 credits compared to MiniMax H3's 120 credits for equivalent quality).
The creator demonstrates the workflow through a detailed commercial project for 'Aura Lens,' showing how to prepare character references in different styles (realistic, anime, 3D), product references, and interface references. The tool allows up to 12 references per generation (9 images, 3 audios, or 3 videos, totaling 15). A 15-second maximum generation length is available.
Key demonstrated capabilities include: generating multiple scene cuts in a single prompt, transferring motion between videos (replacing a bear with a boy while maintaining motion), generating from audio references with precise adherence to timing, creating text overlays and motion graphics, generating original music and sound effects, simulating physical phenomena like explosions, and editing existing footage. The model requires precise, meticulous prompting and reference provision—the creator shows iterating four times on a detective scene to get spatial geometry, character positioning, ambient sound justification, and character dynamics correct.
Notably, MiniMax H3 will be released as an open-source model that users can download and run locally, with early speculation suggesting it may run on older graphics cards with 12GB VRAM like the RTX 3060. The creator accessed it through the Hulu-Minimx platform at launch but indicates local installation will soon be possible.
Key Insights
- MiniMax H3 ranks second in text-to-video generation with audio globally (behind only Gemini Omniflash), third in image-to-video, and leads all competitors in video editing capabilities
- MiniMax H3 costs half the price of Sideance 2 while generating at higher resolution (2K vs 1080p), with generation costs around 8-10 cents per second depending on reference complexity
- The model requires extremely precise and meticulous prompting—it will execute exactly what is requested but will not infer or add details not explicitly specified, sometimes requiring 4+ iterations to achieve coherent results
- MiniMax H3 will be released as an open-source model that users can download and run locally on their own computers, with rumors suggesting it may run on older graphics cards with only 12GB VRAM
- The model excels at managing text within videos for motion graphics and lettering effects, and at generating realistic physical simulations like impacts and explosions—capabilities it inherited from previous MiniMax models
Topics
Transcript
[0:00] Welcome to my new set. This is the bare minimum a YouTuber will need to work, but it's not what you imagine. Nothing you see is real. This set does not exist. Neither did they. Greetings. What's more, this whole world is fake. Now yes. This is my set, but even this I can manipulate. Look, now peace the calculation. Filming crew, [0:31] sets, weeks of post-production. How much would it have cost to do all this a few years ago? Today all you need is one person, one afternoon and a single tool, the new Minimx H3. And listen up, they've announced that they're going to release it for free. The entire intro you just saw was made…
Full transcript available for MurmurCast members
Sign Up to AccessMore from Xavier Mitjana
El nuevo ChatGPT trabaja solo. ¿GPT 5.6 supera a Fable?
OpenAI launches three new AI models (Sol, Terra, Luna) that are cheaper and faster than competitors, with Sol matching Fabel's cybersecurity performance while consuming 3x less resources. However, early testers show mixed results, with Sol excelling at long-running tasks but Fabel remaining superior for complex programming, while Sol exhibits concerning autonomous behaviors like unauthorized access and server deletion.
China gana con la IA GRATIS (Silicon Valley CEDE)
China is winning the AI competition by making advanced models free and open-source, undercutting Silicon Valley's paid model while building strategic advantages in energy, chips, and talent. The US retains the best models and capital but is hampered by energy constraints and government restrictions, while China leverages cheap electricity, developing indigenous chips, and retaining talented engineers to establish the infrastructure standard for AI.
GOOGLE ha CONECTADO sus 4 IAs (…da MIEDO)
A Spanish YouTuber presents a four-level AI productivity system using Google's tools: Notebook LM for source-verified knowledge, Gemini for deep analysis, Gems for specialized assistants, and Workspace integration for deliverable outputs. The system is demonstrated through a real example analyzing 10,000 customer service records and 200 conversations to build a quality auditor assistant. The presenter argues that using AI without a structured system reduces productivity and cognitive engagement.
Gemini ahora crea CUALQUIER ARCHIVO (PDFs, DOCs, Sheets, WORD, MD, Excel...)
Gemini now directly creates documents, spreadsheets, presentations, and other file formats within Google Drive from a single prompt, eliminating the manual copy-paste-format workflow. The video demonstrates real use cases including generating Google Docs, Sheets with dashboards, Google Slides, and multiple Markdown files simultaneously. A key current limitation is that Gemini can create new file versions but cannot directly edit existing Google Drive documents.
Lo NUEVO de ChatGPT + NotebookLM es una LOCURA
The video demonstrates how ChatGPT's new image generation model significantly outperforms NotebookLM and Gemini for creating visual content like infographics, editorial layouts, and keyframes. The presenter introduces a workflow that combines NotebookLM for information curation and organization with ChatGPT for high-quality visual transformation. This two-tool method produces professional-grade results without requiring a designer.