Beneath the giants like ChatGPT, Midjourney, and Gemini, which AI-powered tools are still under the radar yet deserve a closer look? Sane Lebrun, creator of the Upmynt newsletter, shares his picks.
Scientists confirm: This is the most effective way to get your cat’s attention, according to new research
Elderly Couple Refuses Reserved Seats—Viral Train Standoff Sparks Fiery Debate on Courtesy
Sane Lebrun, Founder of Upmynt
Why You Should Never Reheat These Foods in the Microwave – The Hidden Dangers Experts Warn About
I tried the top 5 guard dogs—here’s what makes these breeds the ultimate protectors
Since November 2022, Sane Lebrun has been crafting and publishing the Upmynt newsletter, focused on artificial intelligence and boasting over 5,000 subscribers. Holding a master’s degree from ESCP, Lebrun has previously worked at various tech companies including Gameloft, Decrypt, and most recently Ignite, where he served as Chief Growth Officer.
Bing Video Creator: Free Video Generation Tool
On June 2, Microsoft caught the industry off guard by integrating a video generator into the Bing app, allowing the creation of short, vertical videos stored for 90 days. The surprise came not so much from the functionality itself, but from the choice of technology: the tool uses Sora, developed by OpenAI, previously exclusive to subscribers of the paid ChatGPT Plus service.
This privileged access, made possible by the close relationship between the two companies, comes with limitations. “The features are quite basic: only portrait mode, and videos last up to 5 seconds,” admits Sane Lebrun. “It’s a far cry from what the dedicated Sora generator can do. To create publicly usable videos, extensive iteration is necessary.”
This observation isn’t unique to Bing Video Creator. “Video AI rarely satisfies on the first try,” he reminds us. The tool’s value, accessible with just a Microsoft account, lies elsewhere. It offers a chance to explore the visual possibilities of generative AI, relying on a “fairly robust engine,” he notes. “It should be seen more as a way to experiment with video AI or to quickly generate clips, for example to send to friends,” he concludes.
Duck.ai: Anonymous Interactions with Generative AI
Chatting with conversational agents while staying anonymous and ensuring your data isn’t used to train models or for commercial purposes—that’s the promise of Duck.ai. Launched in March 2025 after two years of beta testing, this tool by the DuckDuckGo search engine adopts a niche stance in the AI sphere, “which hasn’t been known for exemplary data privacy practices or respect for intellectual property,” explains the Upmynt founder.
Completely free, the tool allows you to interact with a chatbot based on various language models, such as GPT-4o mini or Claude 3.5 Haiku. Like most solutions, it responds to questions, makes recommendations, or translates text. Its unique feature? Anonymizing exchanges by replacing your IP address with DuckDuckGo’s, while automatically deleting metadata containing personal information.
In an ecosystem where privacy has clearly never been a priority, especially for free users, the existence of a tool like DuckAI becomes crucial, believes Sane Lebrun. “When you use free AI assistants, the developer reserves the right to use your conversations to train their future AI models,” he emphasizes. “If you input private or confidential data, be aware that it’s processed on servers, often abroad, particularly in North America in the case of ChatGPT, and potentially exploited. Thus, using an AI assistant like Duck AI that anonymizes your exchanges and commits to not exploiting your data, is a godsend for anyone wanting secure and confidential interactions with an AI assistant.”
However, expect less precise responses due to the lack of customization options, such as personalized instructions, or access to the latest models.
The responses are potentially less good than those provided by ChatGPT. But what you come for in Duck AI is the anonymity.
MyLens.ai: Transforming Data into Visual Explanations
Perhaps less known, MyLensAI remains a highly useful tool, especially in a professional context. Why? Because it transforms data or content into customizable, interactive, and explanatory visuals—a capability not often well-managed by generalist solutions. Using a prompt, files provided by the user, or its knowledge corpus, it can produce timelines, charts, or mind maps, all editable through an editing interface.
But does it live up to its promise, especially when faced with complex data? “Yes, the tool is quite impressive in its ability to select a format suited to the requested information. And if not, you still have the option to guide it towards the desired format,” considers Sane Lebrun.
MyLensAI offers a free plan limited to three explanatory visuals which, according to Sane Lebrun, is more than sufficient. “Unless you need to generate explanatory visuals frequently, the free version is quite adequate for regular use,” he concludes.
Reve: Crafting Polished Compositions
Those closely monitoring image generator rankings are familiar with Reve. In March 2025, this under-the-radar player shocked everyone by placing its version 1.0 model at the top of the leaderboard supervised by Artificial Analysis, an entity that provides a neutral evaluation of market technologies, outperforming even OpenAI, which had just replaced DALL-E with 4o Image Generation, known for its adherence to instructions and producing legible text on visuals—a feature still challenging for competitors.
And it’s not surprising, as Reve focuses on the same strengths, and “all at a very low cost,” points out Sane Lebrun. “Reve’s great strength is what you might call its adherence to prompts. It follows your commands to the letter, which means there’s less need to edit a generated image,” he adds. The tool, offering a hundred credits upon sign-up and then 20 credits per day, also excels in text integration. Its limitations are few, though it occasionally struggles to modify a single element in an already generated image. “If you’re looking for visuals that incorporate branding, data, a polished composition, and text, it’s an excellent choice. For purely artistic or creative visuals, Midjourney or 4o might offer more beautiful outcomes or more flexibility in modifications, but at a higher cost and with limitations on text,” concludes Sane Lebrun.
Genspark AI Slides: Generating Ready-to-Use Presentations
This feature left him “falling off [his] chair,” as he shared in his newsletter from April 23. Integrated into Genspark—a platform initially positioned as an AI-powered search engine before evolving into a “super agent” capable of executing tasks autonomously—AI Slides, as the name suggests, generates “ultra-complete and aesthetic presentations,” explains Sane Lebrun.
Specifically, the tool is capable, from a description, of producing a structured, comprehensive presentation on any topic, with relevant material and visuals. It can also enhance the content with up-to-date data, directly collected from the web. The result is quite impressive, and this “from the first version generated,” according to Sane Lebrun. “It’s remarkable to see how, for example, from simple raw data, Genspark can autonomously propose, a dozen slides with in-depth analysis, charts, and associated recommendations. It’s simply astonishing,” he details.
Higgsfield AI: Creating Social Media Videos
Founded by Alex Mashrabov, a former head of generative AI at Snapchat, Higgsfield AI positions itself, since its launch, as a “specialized video effects generator,” recalls Sane Lebrun. Unlike Sora, the tool primarily targets content creators or marketing specialists.
How? By offering a catalog of pre-generated clips in which the user can directly integrate by providing a reference image and, preferably, a detailed prompt. The benefit? Achieving more controlled renders, potentially suitable for social platforms. “The user can only apply a predefined list of video effects fine-tuned in advance by the developer. But this limit also means greater quality and prevents the model from improvising on effects with ultimately poor quality results. On the contrary, every effect offered by the tool is well mastered by the model, and the outcome is thus satisfactory,” details Sane Lebrun.
Higgsfield is perfectly suited for broadcasting content on social media, precisely because the applied effects are well mastered by the model and tailored for creating viral sequences.
Similar Posts
- Adobe Firefly Enhances Video Capabilities: Introduces Veo 3, Sound Effects, and More Predefined Styles
- ChatGPT Unveils New Image Generator: How Good Is It?
- Midjourney Unveils New AI Video Generator: Open to All Users!
- Exciting AI Breakthroughs: 5 Underrated Tools You Need to Watch!
- OpenAI Unveils Sora 2 and a TikTok-Inspired App for Sharing Your Videos
Jordan Park writes in-depth reviews and editorial opinion pieces for Touch Reviews. With a background in UI/UX design, Jordan offers a unique perspective on device usability and user experience across smartphones, tablets, and mobile software.