Hello everyone, as the New Year approaches, today we bring you an incredibly powerful list of AI product recommendations for 2025. This is a recent and meticulously curated collection that includes a variety of highly practical AI tools from both domestic and international sources. Whether you are a student, a working professional, or a creative worker, you will find the perfect fit for you. Don’t wait, take a look now! Note: All products on this list are free versions. If there are any omissions, please add them in the comments section. Thank you!
Firstly, we have domestic AI products that are both free and easy to use (without the need for VPN). Doubao: The ‘beloved child’ of ByteDance, it now boasts a large user base, with the highest app downloads and monthly active users in the country. In December, it added features such as image editing, handwriting, and visual understanding models, making it increasingly powerful. Experience the website at: https://www.doubao.com.
LiblibAI: A platform from Singularity Nebula, boasting powerful features such as SD image generation, model uploading and sharing, ComfyUI workflow, and LoRA training. In December, in collaboration with Star Flow, they launched the new generation image model Star-3 Alpha, which has taken image precision, color representation, and aesthetic expression to a new level. Experience the platform at: www.liblib.art. AI Video Ling 1.6: This version can be considered the global No.1, with comprehensive features including text-to-life, image-to-life, lip-syncing, and motion brushes. Access the platform at: klingai.kuaishou.com. Conch AI: Supports text-to-life, image-to-life, and I2V, with strong generalization capabilities and rich imagination, popular overseas. Experience the platform at: hailuoai.com. Hunyuan Video: A creation of Tencent’s Hunyuan team, currently only supports text-to-life but performs well in command compliance, camera language, image, and dynamic stability. Experience the platform at: https://video.hunyuan.tencent.com. PixVerse V3.5: After the base model update, semantic understanding, motion stability, and detail performance have significantly improved, with new features like start and end frames, and V2V functionality. Generation speed is incredibly fast, capable of creating a 5-second video in just 10 seconds. Experience the platform at: https://app.pixverse.ai. AI Search: Tiangong AI Search has a clear positioning as an AI search engine. It can generate content outlines, mind maps, tables, PPTs, etc., with one click. Advanced search and deep analysis features were launched in November 2024, followed by the Tiangong o1 and 4o versions in January 2025, all free and unlimited. Access the platform at: www.tiangong.cn. Quark: More than just an AI browser, it is a system-level AI that can be summoned on any computer at any time, especially popular among young users. Download the platform at: www.quark.cn. Baidu ‘AI Search’: Ad-free, specially created by Baidu for AI search, with features including AI search, inspiration exploration, intelligent entity responses, and AI creation. Experience the platform at: kimi.moonshot.cn. Wen Xiaoyan APP: In the second half of 2024, Baidu switched it to the search track, positioning it as a new search. The new search mode of search, creation, and chat offers a good experience. Download the app at: https://ima.qq.com. AI Meeting Baidu Netdisk: The listening and recording function is very powerful, able to organize various audio files with one click, and offers a variety of meeting templates for selection.
Recently, the Meeting Assistant feature has been launched, allowing the assistant to join meetings at any time. Experience it at: https://pan.baidu.com/embed/listennote. General Efficiency Tools: The ‘Real-time Record’ assistant in the Tongyi Efficiency section can handle meeting records, meeting audio organization, and learning video organization with ease. Try it out at: https://tongyi.aliyun.com/efficiency. AI Music Dou Bao: Supports generation up to 3 minutes, making it a great choice for those who want to quickly create music clips. Agent Development Platform Kouzi: Easy to use and quick to get started, already supports models like Dou Bao, Moonshot, Tongyi, MiniMax, and can be published to platforms such as Dou Bao, Feishu, WeChat Work, Public Accounts, and Juejin Community. Use the website: www.coze.cn. Intelligent Entities: Here, the focus is on intelligent entities, with only three recommendations: Claude’s Computer Use, Zhi Pu’s AutoGLM family, and Honor Magic phones. Although these products have many shortcomings, they have taken an important step in intelligence. AI Writing Ima Copilot: Tencent’s AI intelligent workbench, capable of searching content from WeChat Public Accounts, Bilibili, and other platforms, helping you organize into a knowledge base, and supporting instant editing, expansion, and polishing. Download at: https://ima.qq.com. AI Programming CodeQwen: Alibaba’s intelligent programming tool, already open-sourced, with 7B parameters, supporting 92 programming languages, and supporting 64K context. Code website: https://modelscope.cn/models/qwen/CodeQwen1.5-7B-Chat/summary. AI Office DingTalk AI and Feishu: Both companies have fully integrated AI into their products, significantly increasing office efficiency. AI Browser Quark: Not just an AI browser, but also a system-level AI. Features include one-click AI search, writing, PPT generation, video summarization, file summarization, screenshot search, and word search. Download link: quark.cn. AI PPT Tools: Kimi PPT Assistant, AiPPT, Baidu Wenku, iFlytek Zhiwen, Tencent Docs, and Qingyan PPT. These AI PPT tools are very user-friendly, especially for PPT beginners, helping you quickly generate high-quality PPTs.
AI Finance Reportify: A convenient tool for investors to research reports, read financial statements, and analyze financial data. Essential for investment professionals. Experience link: https://reportify.cc
AI Translation ‘Marco’: Developed by Alibaba International, supports 15 global mainstream languages, and provides context-based accurate translation. Open for global users. Experience address: www.aidc-ai.com
AI Healthcare Zijing AI Doctor: Developed by a Tsinghua team, 42 AI doctors were launched in December and are currently in the internal testing phase. Experience link: https://www.tairex.cn/agent-hospital, Paper link: https://arxiv.org/pdf/2405.02957v1
AI Product Website AI Toolset: A collection of almost all AI products on the market, with clear categorization and a rich variety of products. Find AI tools here. Experience website: ai-bot.cn
Foreign AI Products, powerful features but require ‘magic’ General Large Models GPT-4o, o1, o3: These models from OpenAI have significantly improved reasoning capabilities, especially o3, which is close to AGI level. We are getting closer to the true era of intelligence. Usage website: www.chatgpt.com
Claude 3.5 Sonnet: The flagship of Anthropic, with particularly excellent text generation effects, surpassing 4o. The visualization function Artifacts is very easy to use. Usage path: Claude.ai
Gemini 2.0: Google’s big move at the end of the year, launching the Gemini 2.0 model, with multimodal and Agent technology as the core, surpassing 1.5 Pro in speed and performance.
API Deployment Model Llama3.3: A new model open-sourced by Meta, with 70B parameters, outperforming Gemini 1.5 Pro, GPT-4o, and Nova Pro in multiple industry benchmark tests. Model download link: https://www.modelscope.cn/models/LLM-Research/Llama-3.3-70B-Instruct
World Model ‘Spatial Intelligence’ Model: Created by the Li Feifei team, capable of generating a realistic 3D world from just one image. Product details: https://www.
worldlabs.ai/blog. System-level AI Copilot: After the September update, it has initiated workflows for Web+Work+Pages, launching products such as Copilot Pages, Office Suite, and AI Copilot.
AI Drawing Midjourney: Updated to version V6.1 on July 31st, priced at $30 per month. It boasts high-quality image output and on August 22nd, it opened its web version to all users, offering 25 free opportunities per day for new users. Visit www.midjourney.com. Recaft V3: In November, it outperformed top models like Flux, Midjourney, and Stable Diffusion in human blind tests. It excels in long text generation and further optimizes body completeness, supporting free canvas. Experience it at https://www.recraft.ai. Flux: An open-source generative model developed by the former SD team, with 12 billion parameters, it stands out in visual quality, image detail, and output diversity. Visit https://blackforestlabs.ai/announcing-flux-1-1-pro-and-the-bfl-api/. Comfy: Similar to liblib in China, it provides users with modular and node-based workflow design based on the SD model. It released a desktop version at the end of October. Join the waiting list at https://www.comfy.org/waitlist. AI Video Veo 2: A Google product, it supports text-to-video and image-to-video, with a maximum resolution of 4k and 8-second videos. It excels in aspects such as lighting, camera, motion, texture, and semantic adherence, surpassing well-known models like Sora, Gen3, and luma. Experience it at https://deepmind.google/technologies/veo/veo-2. Gen 3: Launched by Runway on June 29th, it supports 10-second video generation with cinematic quality and rich lighting details. In August, it launched the image-to-video feature, in September, it introduced lip-syncing and video-to-drawing functions, in October, it released the Act-One feature, and in November, it introduced 3D camera control widgets. Experience it at https://app.runwayml.com/login.Sora officially launched on December 10th, supporting text-to-speech, image-to-speech, and V2V (video-to-video) with a maximum duration of 20 seconds and the highest video quality of 1080P. Experience it at: www.sora.com.
Agent development platform Coze International Edition: A ByteDance product, leveraging ChatGPT, allows for easy creation of AI agents. Use the website: www.coze.com.
GPTs: The official ChatGPT AI agent creator, enabling the creation of personal AI agents based on GPT-3.5/4.0/4o.
AI Search Perplexity: The pioneer of new conversational AI search, requiring a scientific internet connection. On November 19th, the ‘Buy with Pro’ feature was launched, integrating organically with e-commerce. Use the website: www.perplexity.ai.
In addition, products like Felo, New Bing, and SearchGPT are worth trying. Felo, for instance, can search closed platforms like Xiaohongshu and Reddit, and supports multi-model invocation.
Digital human HeyGen: A paid product, priced between $49 to $150 per month. Users simply input a script to generate videos for marketing, communication, sales, learning, etc., within minutes. It supports adding 175 languages and local dialects to videos with AI voice and lip sync. Experience it at: demo.heygen.com.
Voice cloning conch Audio: Clone anyone’s voice in just 30 seconds, supporting 12 languages. However, this feature is currently only available overseas and requires a scientific internet connection. Experience it at: https://www.hailuo.ai/audio.
AI Music Suno 4: Can write songs with one click, but requires lyrics creation through other AI first. Currently free, but requires a scientific internet connection. Suno has been upgraded to version 5, capable of generating songs up to 4 minutes long, with significant improvements in audio quality, lyric creation, and music structure. Use the website: suno.com.
AI Writing Napkin: An AI writing assistant tool founded by former Google members, capable of visualizing text in various structures, including flowcharts, diagrams, and tree diagrams. Experience it at: https://app.napkin.ai/signin.
AI Scientist: A new model launched by former Google researchers, claiming to be the world’s first fully automated scientific research AI system, capable of the entire scientific process from conception to paper writing and peer review.
Users can generate a research paper with just an idea, costing less than $15. The open-source address is: http://github.com/SakanaAI/AI-Scientist, and the research paper link is: https://arxiv.org/abs/2408.06292.
AI in pharmaceuticals, AlphaFold 3, developed by Google DeepMind, can accurately predict the structures of proteins, DNA, RNA, and ligands, among other biomolecules. In October, the developers of AlphaFold received the Nobel Prize in Chemistry, and in November, Google made the model open-source for the benefit of humanity. The model’s website is: https://golgi.sandbox.google.com/about. AI PPT tool, Gamma, does not require a VPN and can be registered with a QQ email. It boasts high-quality layout and, arguably the pinnacle of AI PPT tools. However, it is a paid service, with a personal version costing $10 per month. The website is: https://gamma.app. AI coding assistant, Cursor, is based on large language models like ChatGPT and Claude, supporting multiple languages such as Python, Java, C#, and Javascript. It offers features like code auto-generation, editing, completion, optimization, and explanation. It became particularly popular among programmers in December, with many using it as an alternative to VS Code. The experience address is: https://www.cursor.com. AI gaming, Oasis, is the world’s first real-time generated AI game, developed jointly by two startups, Etched and Decart AI. Every frame you experience is from the real-time prediction of diffusion models, with the game graphics continuously rendered at a rate of 20 frames per second, with zero latency. The experience address is: https://oasis-model.github.io/.
