MiniMax与fal直播解析H3：覆盖多模态创作、原生音频与LoRA工作流

## Common questions about MiniMax H3 MiniMax H3 is an open-weights, general-purpose multimodal video model. Instead of a separate model for each task, MiniMax H3 reads text, images, video, and audio in one unified context and generates coherent audiovisual results from any mix of them. It supports text-to-video, first-and-last-frame, reference-to-video, and precise video editing. MiniMax H3 is released with open weights, so it is an open foundation you can explore, customize, and build on rather than a closed endpoint. fal is a Day 0 ecosystem partner, which means you can call the hosted MiniMax H3 API on fal.ai from launch without provisioning GPUs, and still have the option to work with the weights directly for your own research and fine-tuning. [...] ### Sound Composed to Picture Every generation returns native stereo audio: original score, dialogue, foley, and room tone timed to the cut. Give MiniMax H3 a reference recording and it will transfer or clone that voice onto your cha

## What can you create with MiniMax H3? H3 is designed for creative briefs that combine several kinds of reference material and require video, sound, text, motion, and brand details to work together. 01 ### Ads and ecommerce videos Create product reveals, vertical social ads, brand films, animated posters, and campaign concepts with stronger text and product-detail rendering. 02 ### Reference-led stories and editing Transfer motion, preserve a subject, follow visual or audio references, regenerate scenes, and describe complex edit relationships in natural language. 03 ### Open-weight research and deployment Run the H3-Base checkpoints, study the architecture, build custom inference workflows, or fine-tune the released model for specialized creative tasks. [...] Unlike video systems divided into separate text-to-video, image-to-video, motion-reference, subject-reference, audio, and editing models, H3 is designed to express those tasks through natural-language instructions insi

Skip to content GENERATIVE AI # fal Adds LoRA Training for MiniMax H3, Starting With an Open-Source Realism People LoRA VP Land Aug 11, 2026· 2 min read Custom LoRA training for MiniMax H3 is now live on fal, and the company demonstrated it with Realism People, an open-source LoRA it says pushes the model toward photorealistic humans. The trainer targets MiniMax H3. It fine-tunes the open-weight model that generates video, image, audio, and text. Realism People ships open source. fal posted the LoRA weights to Hugging Face, tuned for skin, eyes, and motion. The trainer covers several input modes. fal points to entry points for text-to-video, image-to-video, first-last-frame, and reference-to-video. More LoRAs are coming. fal says additional LoRAs are on the way. [...] WorldClaw, a new paper from Tencent Hunyuan, describes a text-to-3D system that builds large, explorable scenes and returns editable, instance-level assets rather than a single locked render. The method leans on

FN1. How you doing? Good to have you on the stream. If you are just joining us, we are building a music video and uh we're doing it live and it's scary and fun and [laughter] messy, really messy, but it's looking good. It's coming out good. So, okay. Going to get this prompt in here and see how this works. [snorts] All right. So image one is the singer, image two is the location. Not sure. I guess I got to reupload my background answers because they're not showing up. [music] Oh. Oh, are you kidding me? [laughter] It's giving us an reference audio issue, saying that the reference audio is longer than 15 seconds. I don't know how that could be. Um, let me check this. [music] All right, let me get the full details here. Did I put the wrong audio on perhaps? [sighs] Okay, well, let me just [...] I don't know. All right. Be serious, man. Come on. Shave an eyebrow. I need shaving an eyebrow. Put like a little line through it or something. Shave an eyebrow. [sighs] All right. What am I doing

Join @MiniMax_AI × @fal for a deep dive into building with MiniMax H3—from multimodal references, native audio, and editing to LoRAs, model

Join @MiniMax_AI × @fal for a deep dive into building with MiniMax H3—from multimodal references, native audio, and editing to LoRAs, model