New AI video editor, new AI 3D models, new open-source AI beats DeepSeek, new VEO competitor

New AI video editor, new AI 3D models, new open-source AI beats DeepSeek, new VEO competitor

by AI Search

Applied Sciences & Engineering AI — News & Applications UYComputer scienceUYQArtificial intelligence (AI)
🎙 AI Search 👥 727K 📅 June 22, 2025 ⏱ 35 min 👁 108K 📄 news review 🧭 2026-09-07
Available in: English (current) Français

Keywords

TAIRHunyuan 3D 2.1PartTrackerLoraEditMinimax-M1

Summary

This video from the channel AI Search provides a weekly roundup of recent AI developments, covering a range of tools and models. The host begins with TAIR, an image restoration model that excels at sharpening blurry text. Next, he showcases Hunyuan 3D 2.1, an open-source 3D model generator from Tencent, demonstrating its ability to create detailed 3D assets from single images. He then introduces PartTracker by NVIDIA, which segments 3D models into editable parts, and LoraEdit, a video editing technique that applies edits from a modified first frame to the entire video. The video also covers Minimax-M1, a large open-source language model with a 1 million token context window, and ImmerseGen, a tool for generating realistic 3D virtual environments. Finally, he reviews Midjourney V1, a new video generation model, and briefly mentions other tools like InterActHuman and POLARIS. Throughout, the host provides practical demonstrations and links to official sources.

149 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video offers substantial value by aggregating multiple AI releases in a single update, saving viewers time. The host demonstrates each tool with real examples, which helps illustrate their capabilities and limitations. The argumentation is largely based on the tools’ own performance in demos and benchmark scores provided by the developers. While the host does not critically evaluate the methodologies behind these benchmarks, he does provide context by comparing with existing alternatives. The reasoning is generally sound, though it relies heavily on the claims of the tool creators.

Scientific Rigor, Source Quality, Title Accuracy

The video maintains a high level of scientific rigor by consistently linking to official project pages, GitHub repositories, and research papers for each tool mentioned. The host also provides a clear structure with timestamps, making it easy to navigate. The title accurately reflects the content, which is a news review of recent AI developments. The sources are credible and directly relevant to the tools discussed. The video does not include any obvious misinformation, and the host is transparent about the limitations of some tools, such as Midjourney V1’s lack of text-to-video and low resolution.

197 words

Title / Content Match

The title accurately reflects the content, which covers new AI video editors, 3D models, an open-source AI beating DeepSeek, and a VEO competitor.

Quality & Reliability

8/10

The video provides a comprehensive overview of recent AI releases, with direct links to official sources and demonstrations. The information is presented clearly and appears accurate, though some claims (e.g., benchmark superiority) are based on the creators' own reports.

Chapters

Cited Sources

  • TAIR: Text-Aware Image Restoration — Official project page for the text-aware image restoration model.
  • Hunyuan3D 2.1 — GitHub repository for Tencent's open-source 3D model generator.
  • PartTracker — NVIDIA research page for the 3D part segmentation tool.
  • LoraEdit — Project page for the video editing technique using first-frame edits.
  • Minimax-M1 — GitHub repository for the open-source large language model.
  • ImmerseGen — Project page for the 3D world generation model.
  • Midjourney V1 Video Model — Official announcement of Midjourney's first video generation model.
  • InterActHuman — Project page for the human interaction generation model.
  • POLARIS — GitHub repository for the open-source AI model that beats DeepSeek.
  • AlignYourFlow — NVIDIA research page for the diffusion model alignment technique.

Concurring Sources

  • Hailuo 02 Review — The host's own review of the Hailuo 02 video model, which is referenced for comparison.

External References

Contribution & Novelties

The video provides a timely and comprehensive overview of recent AI releases, highlighting several novel tools and models. The main novelty lies in the aggregation and practical demonstration of these tools, which helps viewers understand their potential applications. The video also emphasizes the trend of open-source AI models, such as Minimax-M1, which offer competitive performance at lower costs.

Pour aller plus loin :

  • Diffusion Transformers — The underlying architecture for many image and video generation models, including TAIR.
  • Mixture of Experts — The architecture used by Minimax-M1, which combines multiple expert models.
  • Text-to-3D Generation — The field of generating 3D models from text or images, as demonstrated by Hunyuan 3D 2.1.

111 words

Radar Profile

The radar profile shows high scores in information quantity and quality, reflecting the video's comprehensive coverage and use of credible sources. The technical level is moderate, making it accessible to a broad audience while still providing useful details. The overall reliability is high, with no significant discrepancies noted.

Reliability 8/10

💬 Positif. Sur les 30 commentaires analysés, le climat est très positif, avec des éloges pour la couverture complète de l'actualité IA et la qualité des démonstrations. Plusieurs commentaires expriment leur appréciation pour la diversité des sujets traités et la fréquence des mises à jour.