Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

MiniMax H3

AI overview

Sign in and the AI will write an overview from our coverage.

Headlines · 19

  1. FreeVideo v0.5.0 adds local H3 image generation

    FreeVideo v0.5.0 adds local MiniMax H3 image generation for images up to 16 megapixels, according to the project’s release notes. The open-source local inference engine runs as a ComfyUI plugin and supports text prompts, first and last frames, and image, video and audio references. The project says FreeVideo can run MiniMax H3 on consumer GPUs with as little as 8 GB of VRAM and 16 GB of RAM. Its code is licensed under Apache License 2.0, while the model weights use the MiniMax H3 Community License, which includes territorial and acceptable-use restrictions.

    GitHub Trending · Python · 🔥 3
  2. Suno launches Visuals for AI-generated music videos

    Suno has launched Visuals, a feature for turning songs into AI-generated images, video clips of up to 15 seconds and lyric videos with optional lip synchronization. Users can start from prompts, uploaded references or Suno presets, and edit clips on a timeline. Suno says the feature runs on H3 Move, a customized and safety-filtered version of MiniMax’s open-source H3 video model, and uses separate visual credits. Web availability is confirmed, but reports differ on mobile access: some say iOS and Android are available, while another says the mobile apps were due in the following weeks.

    suno (X) · 🔥 6
  3. AI Video Tools Enter a Double 11 Price War

    36氪 人工智能, citing 叁拾代, reports that Double 11 promotions have intensified price competition among AI video tools. LibTV advertised up to 30 days of free Wan3.0, Kling4.0 and Minimax H3 usage, while 即梦AI and AniShort offered subscription and first-purchase discounts. The report says the market is shifting from feature parity toward creator ecosystems, vertical workflows and new business models as products become increasingly homogeneous.

    36氪 人工智能 · 🔥 0
  4. MiniMax H3 Image gains native ComfyUI support

    Comfy Blog reports that MiniMax H3 Image is now supported natively in ComfyUI, with open weights available for local GPU execution or API access through Partner Nodes. The single checkpoint handles text-to-image generation and reference-image editing, including multi-reference workflows with numbered image inputs. Comfy Blog also says the CFG-distilled model uses six fixed sampling steps at guidance 1 and supports native output up to 4096×4096 and larger aspect-ratio resolutions.

    Comfy Blog · 🔥 0
  5. Utopai Studios appoints Hans Ku as chief product officer

    Utopai Studios announced that former Disney AI executive Hans Ku has joined as chief product officer, reporting to CEO Cecilia Shen. He will oversee product, engineering, research and business development, with a focus on expanding PAI across original productions and industry partnerships. The reports also say he previously helped Disney's collaboration with OpenAI around Sora.

    品玩 实时要闻 · 🔥 0
  6. Pika releases open-weight Pika H3 Flex for six-step audio-video generation

    Pika released Pika H3 Flex, an open-weight derivative of MiniMax H3, with LoRA adapters, inference code, base components, local tools, and integrations for ComfyUI and Hugging Face Diffusers. It supports six-step joint audio-video generation from text, first-and-last-frame, and multimodal-reference inputs, with native 480P or 768P output of roughly 5–15 seconds and synchronized stereo audio.

    品玩 实时要闻 · 🔥 0
  7. Utopai Studios Raises $60M at Over $1B Valuation

    Utopai Studios reportedly completed a $60 million funding round in October 2026 at a valuation above $1 billion, citing Variety, Dealroom and The Information. The AI-native film studio operates the customized Utopai X video model and PAI production-intelligence system across theatrical distribution and IP operations. The reports also say it received indications of interest for a further $2–3 billion round.

    InfoQ 中文 AI&大模型 · 🔥 0
  8. Open-H3-Regenerate-2K Released for 2K and 4K Video Regeneration

    NVIDIA and MIT researchers on the Efficient-Large-Model team released the Open-H3-Regenerate-2K weights on Hugging Face and inference code in the NVlabs/Sana repository. Based on MiniMax H3, the deployable implementation regenerates low-resolution drafts at 2560×1440 and supports up to 3840×2160 native 4K output. The team reports that a prewarmed five-second 2K processing pass on one NVIDIA B200 GPU fell from 87.26 seconds to 22.45 seconds, while its UniPercept score reached 60.0 versus 53.4 for the official API.

    品玩 实时要闻 · 🔥 0
  9. MiniMax H3 Tops SuperCLUE’s Updated Image-to-Video Leaderboard

    SuperCLUE updated its image-to-video model leaderboard, placing MiniMax H3 first in the formal domestic ranking with an overall score of 71.46 and an application-capability score of 75.82. Alibaba’s Wan3.0 ranked second with 66.63, while ByteDance’s Seedance2.5 and Kuaishou’s Kling V3Pro tied for third under the rule that scores within one point are considered equal. Google’s Gemini Omni1.1Flash and xAI’s Grok Imagine Video1.5 were reference models outside the formal domestic ranking, and some tested models received zero scores after triggering constraints involving real-person appearances, generation duration or content review.

    AIbase AI新闻 · 🔥 0
  10. MiniMax H3 community update adds faster generation and runtime integrations

    MiniMax reports that community contributions to its open-weight MiniMax H3 video-generation ecosystem include 8-step reference-conditioned video and audio, 4-step distillation, sparse attention, and reference-token compression. The update also adds serving and workflow integrations for GPU servers, TPUs, and Apple Silicon, bringing the H3 ecosystem index to 11 additions. MiniMax says these changes can make experimentation more affordable, creative control finer, and deployment choices broader, though those benefits remain vendor claims.

    MiniMax_AI (X) · 🔥 0
  11. AI Reshapes Content Production Workflows

    TMTPost reports that multimodal models such as Seedance and MiniMax H3 are being integrated into content-production workflows for advertising, gaming, e-commerce and interactive media. The report says growing demand for channel-specific and frequently refreshed assets is outpacing conventional production, making AI adoption increasingly important for repetitive, high-volume tasks. It describes an unnamed large game company using a self-built technology platform and MiniMax H3 and Seedance to produce marketing images and videos, reducing some production times and costs. MiniMax H3 was also used in Bilibili’s updream creator platform, while other platforms such as Workbuddy and Miora were adjusting their tools for new models. The report argues that practical adoption depends on reliability, controllable costs, creative consistency, lower review and rework rates, and campaign feedback—not technical benchmarks alone. It also describes interactive film-game formats as an emerging direction whose sustainable commercial cycle has not yet been established.

    钛媒体 · 🔥 0
  12. Utopai opens MIPCOM as AI-native studio; Banderas to star in Cortés

    On October 13, AI-native film studio Utopai took the opening keynote stage at MIPCOM in Cannes, pitching a studio model built around original IP, human creativity and its own film infrastructure, and disclosing that a partnership with Korea's CJ ENM Studios is being advanced. The same day, Antonio Banderas announced he will star in Utopai's film Cortés, which the report describes as his first theatrical feature with AI involvement. The report also says Utopai plans three theatrical releases in 2027 — Cortés, Half Moon and The Most Serious Fart — and that its custom video model Utopai X ranks second worldwide with a score of 1149 on Artificial Analysis's blind text-to-video-with-audio leaderboard.

    钛媒体 · 🔥 0
  13. Vibe Video: code-rendered clips squeeze Kling 4.0 on two fronts

    After Anthropic released Claude Opus 5.5, creators began using it — with Playwright and FFmpeg — to turn code-written HTML, Canvas, Three.js and GSAP animations into finished videos, a workflow the article calls "Vibe Video." Its strengths are deterministic, addressable frames and low cost (Opus 5.5 API at $4 per million input tokens and $20 per million output tokens), but it cannot render photoreal faces or simulate physics and cannot take reference video as input. On the professional side, Seedance 2.5 offers native 30-second clips and up to 50 multimodal references, Kuaishou's Kling 4.0 added 10-bit HDR, up to 15 reference assets and up to 10 keyframes in late September, and MiniMax H3 is a 33B open multimodal model at 2K/24fps with synchronized audio and video. The article argues general LLMs are capturing the planning stage while dedicated models become rendering engines, leaving Kling 4.0 squeezed between Seedance 2.5 and open-source H3.

    36氪 人工智能 · 🔥 0
  14. Utopai X ranks second on AI video leaderboard as studio builds AI production stack

    An InfoQ deep dive profiles Utopai Studios, a Mountain View-based AI film company valued at about $1 billion. Its video model Utopai X, post-trained on top of MiniMax H3, ranks second globally on Artificial Analysis' text-to-video-with-audio leaderboard at roughly 1150 Elo (behind Wan 3.0) and is listed first for audio synchronization and physics. The company also ships the PAI production platform, which Japan's DeNA has begun using in animation workflows, and holds more than 20 film projects, six of which are in production or pre-production.

    InfoQ 中文 AI&大模型 · 🔥 0
  15. MiniMax H3 claims physics performance near closed-source SOTA

    MiniMax promoted its open-source video world model MiniMax H3 on X, saying it is "almost on par with closed-source sota" on physics. The post cites an Einsia evaluation that gave eight state-of-the-art video world models a physics exam called "World Models' Last Exam in Physics," where the top score was 57.76/100 and the average composite score was 26.61/100; it also notes that a free-fall video passed the consistency screen. The near-SOTA claim is a vendor statement, and the post does not provide independent verification or full methodology.

    MiniMax_AI (X) · 🔥 0
  16. MiniMax H3 live video generation on a single RTX 5090

    A Comfy blog post presents ComfyStreamerH3, an open-source custom ComfyUI node that runs MiniMax H3 on one 32GB RTX 5090, producing 15 seconds of 448×256 video in 15 seconds or less with a four-step setup. The listed optimizations include sparse attention that applies full attention to only about 20% of video blocks, NicoLab28's ClipProj adapting Qwen3-VL-4B as a stand-in for H3's 32B text encoder (about 4.5GB vs 15.7GB), FastVideo's pruned INT8 checkpoint, a fused FP4 MLP, and Kijai's INT8 tiled decoder overlapped with NVENC encoding, cutting VRAM needs from 80GB to under 30GB. The author says running such models continuously can cost upwards of $6,000 per day while RTX 5090 rentals start near 70 cents an hour, and notes FastVideo's Preview V1 took 47.2 seconds for a 15-second 1344×768 clip on one B200. The clips still show visible artifacts, and the author suggests passing them through a cheap upscaler such as SolRefinery.

    Comfy Blog · 🔥 0
  17. Utopai X's No.2 video ranking questioned: post-training on MiniMax H3 under scrutiny

    On Sept 30, 2026, Artificial Analysis's text-to-video blind-voting leaderboard placed Utopai Studios' Utopai X second worldwide with 1150 points, behind only Wan 3.0 and the top-ranked US model, prompting a wave of Chinese-language coverage calling it an "AI video dark horse" with a "proprietary custom model." The article says roughly 90% of that coverage ignored the leaderboard footnote "Utopai X (based on MiniMax H3)" — it is not trained from scratch but post-trained on MiniMax's open-source base with film-production data. It raises three questions: cost-effectiveness (base scored 1136 vs 1150, and no training compute or inference configuration has been disclosed), whether the fine-tuning can migrate to a new base model, and whether its iteration speed can match native models such as Seedance's continuous data flywheel.

    钛媒体 · 🔥 0
  18. Utopai X Reportedly Ranks Second in Artificial Analysis Video Test

    Utopai Studios’ Utopai X reportedly scored 1,150 (±10) in an Artificial Analysis blind video evaluation, ranking second globally and first in the United States. The article says the model was post-trained from MiniMax H3 for film production and led the evaluation’s audio-sync and physics categories, but these claims are presented through a promotional report and are not independently verified here.

    新智元 · 🔥 0
  19. Utopai X reportedly ranks second in Artificial Analysis Video Arena

    A Machine Intelligence report says Utopai X ranked second globally, with an Elo score of 1150, in Artificial Analysis's Video Arena in late September. The article attributes the result to custom post-training built on the MiniMax H3 architecture and integration with Utopai's Production Intelligence platform, while noting that the claims come from the reported evaluation and company materials.

    机器之心 · 🔥 0

Experience and discussion from the community

Share my MiniMax H3 experienceAsk about MiniMax H3

FreeVideo: A local inference engine for generating MiniMax H3 videos on consumer GPUs.

What problem it solves FreeVideo enables local video generation on consumer hardware while coordinating VRAM, system memory, disk, weight placement, compute precision, and attention kernels. It provides a ComfyUI workflow and launchers that simplify setup across Windows, macOS, and Linux. Key features Hardware-adaptive execution — The engine selects GPU-specific weight formats and probes available…

github.comhttps://github.com/FlashML-org/FreeVideo