PH热榜 | 2026-03-02

Hey Product Hunt 👋

I’m Romàn, co-founder of GojiberryAI

We built Gojiberry because outbound is broken.

Founders and small sales teams waste hours:

• Scraping random leads

• Sending generic “Hey {{first_name}}” messages

• Guessing who might be interested

• Burning accounts with bad automation

And the worst part? Most of those people were never ready to buy.

So instead of automating spam… we decided to automate intent.

GojiberryAI is an AI GTM Brain that:

→ Detects high-intent buying signals (profile views, job changes, funding, competitor engagement, content interactions)

→ Enriches and qualifies leads automatically

→ Generates hyper-personalized LinkedIn conversations

→ Centralizes everything in one inbox

→ Lets you fine-tune campaigns with an AI Co-Pilot

Unlike traditional outreach tools that focus on volume, we focus on signal first, message second.

The result:

• Higher reply rates

• More booked demos

• Less manual work

• No copy-paste templates

We’ve used this system to generate hundreds of conversations and scale from 0 to $1M ARR, and now we’re opening it to the PH community.

🎁 Product Hunt Special

We’re offering an exclusive launch discount for the PH community.

Use code: PH10 for 10% off your first month.

We’ll be here all day answering questions, sharing our exact stack, and being fully transparent about what works (and what doesn’t).

Thanks for checking us out, excited to hear your feedback 🚀

These guys are the best at solving a problem that nobody solved well, at a right price in that space AND they build an integration with Breakcold :)

Go try it! GG @roman_cz @pierre_eliott_llt @dylan_teixeira

Nice app and a very serious team 💪

another day another gold!

Been seeing yall a lot on socials. very well cracked the game! All love and support :) Would love to learn a thing or two :) @roman_cz @pierre_eliott_llt @dylan_teixeira

amazing product, made by an amazing team

LFG

Saw this launch as soon as thought expanding my product for b2b, gonna try it! congrats on the launch

Love the idea of automating intent instead of spam. Feels like the right evolution for modern sales teams 🚀

Congrats for the launch!

This is solving a real pain point. Cold outreach has always felt like shouting into a void, the intent signal approach makes so much more sense.

Curious about one thing, how do you handle signal accuracy? Like if someone liked a LinkedIn post about "sales tools," does Gojiberry flag them as a buyer, or is there more filtering happening behind the scenes to reduce false positives?

Also the "one prompt to build a lead list" UX sounds really clean. Is that powered by a custom model or GPT-based?

Congrats on the launch, B2B teams are going to love this!

Congrats on the launch! Does the AI handle follow-ups if a lead doesn't book a demo immediately?

Good luck with the launch!

How does it actually detect high-intent buying signals? Curious to use it. Congratulations on the launch, @roman_cz!

Good luck guys! The tool looks solid. Surely research and try it on weekend.

I was an early customer (and still am a customer), and this tool has always been worth keeping, always gave me good results. The team is wonderful too, very helpful guys.

@roman_cz @pierre_eliott_llt What an amazing team and product. Have been using gojiberryAI for a while and can confirm its a game changer. Plus they are always available to assist whenever I needed help.

Congrats on the product and the launch!

QQ: The "intent first, message second" approach makes sense, but how do you filter false positives? Like, someone liking a competitor's post could mean they're evaluating options, or it could just mean they thought the post was interesting. How do you weigh different signals to avoid chasing people who aren't actually in buying mode?

Going to check it our now! I started building a similar solution, but this is much better :)

"Finally! Finally, there's a tool that distinguishes between quantity and quality in sales. What is interesting most about Gojiberry is focusing on buying intent instead of spamming people with random messages. In a world where LinkedIn is crowded with annoying messages, having a tool that understands signals (like job changes or competitor engagement) and crafts a personalized message based on them—this is the real future of sales. A question for the team: Does the tool currently support Arabic in signal detection or message writing? Cheers to you on the fantastic launch and on to the next million 🚀

@dylan_teixeira

How do you promote B2B SaaS? I am currently launching my AI-based B2B product on the market, and I am really interested in where you get your customers from.

This is super interesting focusing on intent over volume makes a lot of sense

Curious — with tracking signals from platforms like LinkedIn and automating outreach, how are you handling things like data privacy and account safety?

Especially avoiding issues like platform restrictions or unintended access risks

Great idea! Is there pricing for solo founders or startups? $100/mo is steep for most people.

High reply rates are good, but pipeline quality is what matters. Have you seen improvements in close rates too, or mainly top-of-funnel lift?

Congrats guys! With your experience from gojiberry, what’s one GTM lesson you’ve learned? And what’s the inspiration behind the name “Gojiberry”?! It sounds sweet, I like it! lol

Hey @roman_cz congrats on your launch and great product! Does this primarily only work with LinkedIn and would this require Sales Navigator enabled to enable DMs to people with intent but not yet in your network?

Hey Product Hunt! I built crawler.sh because I kept running into the same problem: every time I needed to audit a website's SEO or extract its content, I had to choose between bloated enterprise tools, slow cloud services, or stitching together a bunch of scripts.

crawler.sh is a single tool that does all three: crawling, SEO analysis, and content extraction, from the terminal or a native desktop app. It's built in Rust so it's fast, it runs locally so your data stays private, and it outputs standard formats so you can take the results anywhere.

Local-first approach is underrated, most SEO tools are cloud-heavy and slow. The fact that this runs from terminal AND has a native desktop app is a nice touch for different types of users.

The Markdown extraction feature caught my eye specifically,been looking for something that can cleanly pull content without all the HTML noise. How does it handle JavaScript-heavy sites like React or Next.js? Does it wait for JS to render or purely static crawl?

Also curious, AEO (Answer Engine Optimization) is still pretty new territory. What kind of checks are you running for that specifically? Most tools haven't caught up there yet.

Congrats on the launch, terminal-based dev tools don't get enough love on PH!

Man, markdown extraction is such a lifesaver when dealing with giant websites. Way easier than my old 'copy-paste marathon' method. 😂

Loving the ease of this. Is there a team plan?

Great tool for SEO enthusiasts! How does Crawler.sh handle dynamic JavaScript-heavy websites during the crawl?

Congratulations on the launch, @mehmetkose! Curious to see how the platform provides feedback on the issues

Well done and congrats for the launch!

Fast, private, and outputs standard formats, that’s how tools should be. Congrats on the launch 🙌

I really like this local-first approach. Congrats!

Does this handle JavaScript-heavy sites (React, Next.js)? Or is it pure HTML crawling? I believe that's usually where terminal crawlers fall short vs the bloated enterprise tools, but I also might be mistaken.

How is that different from ScreamingFrog?

Love the local-first + Rust combo 🔥 Fast, private, no bloated SaaS dashboards.

Markdown extraction is a big win for LLM workflows.

Quick one does it render JS-heavy sites (React/Next), or is it mostly static crawling?

Love this. A fast local tool that actually respects privacy and keeps things simple is refreshing, especially for people who just want clean data without the overhead. Congrats on the launch, this feels built by someone who’s actually done the work.

Love the positioning, local, fast, no bloat is refreshing. How deep does the AEO analysis go? Is it just schema and structured data checks, or does it also evaluate content for AI-answer readiness?

Love the local-first approach especially keeping everything fast and private

Curious — since it can crawl and extract entire sites, how are you handling safeguards around accidentally pulling sensitive or restricted data?

Feels like that could be important as usage scales

one click deploy + runs 24/7 is the dream. whats the average monthly cost to keep it running?

Hi everyone!

Kimi Claw makes running @OpenClaw super easy — one click and your personal assistant is deployed to the cloud, online 24/7, pre-loaded with K2.5.

Have been running my local Claw full-time with @Kimi AI - Now with K2.5 for weeks now. It’s one of the strongest and most cost-effective models for long-running agents, solid reasoning and tool use at a very reasonable token cost. The token consumption tells everything:

If you don’t want to self-host, deal with local security permissions or keep a machine alive 24/7, this official version from Kimi is by far the most hassle-free way. You get built-in personality, long-term memory, 5,000+ ClawHub skills, and seamless Telegram integration.

Open for Allegretto members and above. Try it here: https://kimi.com/bot

What are the benefits compared to running it with kimi API locally? Except of easier deployment

For everyday productivity, what tasks does K2.5 outperform traditional AI chat tools at?

This looks super powerful especially the always-on agents with long-term memory

Curious — with persistent memory and 24/7 execution, how are you handling security around stored data and permissions?

Especially making sure agents don’t take unintended actions or expose sensitive info over time

Lol I need this so my boss doesn't see me endlessly doomscrolling on X

So bascially command + F3 on mac with a document open behind your browser?

Love that it kills audio too — that's the detail most "boss key" tools miss. Alan's right though, the cover doc should probably be a terminal with Claude running these days 😄

So simple, so good. This would be nice for if I got one of those unlucky pops up while looking for sports streams. Nice job.

Love how simple and practical this is 😄

Curious — since it’s hiding all active windows and muting everything instantly, are there any edge cases where certain apps (like screen recordings or notifications) might still leak through?

Feels like those small things could break the illusion in real scenarios

Ah, you've triggered some ancient memories here. We used to call these "Boss key" to hide whatever game you're playing when the manager walks into the office.

Obviously, it's less needed these days with remote working from home, a multi-tasking OS that can change apps quickly, we can all very easily switch away from "Leisure Suit Larry in the Land of the Lounge Lizards" - or whatever other important non-work activity we're hiding.

Might be better making it look like a terminal running Claude or Copilot these days though :)

Are there still workplaces that actually justify this? 😅

Love the icon! Cute kitty!

And ofcourse, feels damn useful too.

A very simple concept that offers an answer to a very old problem. I might not necessarily be the target group for this, but this still made me think of the many times I was caught red-handed doing something unrelated to work/study.

The genius of this lies in its simplicity, which is very hard to reach. (Might need to get used to this particular hotkey, but it will soon be a part of our muscle memory.)

Incredible light app/widget. I hope to put it to good use in the near future.

P.S. : I honestly think one additional update to this could be the option to enable a particular software application to boot up instead of a particular document. I believe there is scope to extend this app's use case towards general productivity by offering a trigger to instantly switch back to high priority tasks.

Shopping? Reddit? *pfft not me 🙈lol I really like the new Guard Mode — it would be so awkward having all visuals hidden but that YouTube audio still playing.

I never expected to get grilled and testify in billion-dollar lawsuits, but there I was. The topics: privacy, consent, biometric data. I’ll spare you the details, but honestly it’s less about those lawsuits and more about the experiences that led up to them.

What I came to understand is this: around the world, your public identity — how you show up online — can impact your life, your freedom, and your finances in ways you’d never expect. And the wave of AI tools has made the cost of manipulating someone’s identity effectively zero.

My co-founder, Tal, and I saw this wave coming years ago and knew we had to do something about it.

WEIR AI is a privacy and consent-centric identity rights platform. We help people track their identity online, to protect it or earn from it.

Building it meant reinventing identity recognition technology from scratch, purpose-built to put people in control without limiting their commercial options.

Nothing off the shelf could do what we needed, so we built it ourselves. We formed WEIR as a Public Benefit Corporation with a mission rooted in the privacy, safety, and security of people and institutions. For us, that was never a question.

We’ve been grinding for a while and documenting some of it on social media.

The good news? Our instincts turned out to be dead on.

The less-than-good news? It’s taken much more effort, resources, and time than we ever planned for.

But we’re here now, and we’re genuinely excited to show you what we’ve been working on

We’ve learned from our design partners and customers about their needs and how we can help. We have a lot more ahead of us, but we’re proud of what we’ve built and how it’s already impacting people’s lives.

Whether you’re the biggest celebrity in the world or a regular Joe, WEIR AI helps you protect, control, and benefit from what’s uniquely yours — your public identity.

Hey Product Hunt.

I'm Tal, co-founder and CTO of WEIR AI.

I've spent twenty years building face recognition systems, first in academia, then leading face recognition development at AWS and Meta. I know how this technology works, and I know what it makes possible, for better and for worse.
Here's the problem: your face shows up in places you never agreed to. Social media posts, ad campaigns, AI-generated content. Sometimes you don't even know it's happening. And right now, there is very little you can do about it.

Why? Because the technology that could help, face recognition, has become so legally toxic that no large company will touch it. Billions in fines and a tightening web of biometric privacy laws mean that the companies with the resources and expertise to protect you have every incentive to stay far away. So the tools that could give you control over your own face simply don't exist. Meanwhile, the bad actors using your likeness without permission don't care about any of that.

We built WEIR AI to close that gap. Our technology was designed from scratch to be privacy-preserving at the algorithmic level, not as a legal patch on top of old systems. That's what makes it possible to do what nobody else will: find where your face appears and put you in control of it.

You decide the rules: take something down, require attribution, or get paid when someone uses your face commercially. Your face, your call.

For creators, public figures, and professionals whose appearance is part of their livelihood, this is not a hypothetical problem. It is happening to them right now, every day.

My co-founder Gary McCoy and I started WEIR AI as a public benefit corporation because we believe the people who built the technology that makes face recognition possible (yes, that includes me) have a responsibility to also build the tools that put control back in people's hands.

We call it Public Identity Management. Try it and let us know what you think.

Tal and the WEIR AI team

Tal's framing of the core paradox is what makes this compelling: the technology capable of protecting you has become so legally toxic that the companies who built it won't touch it. Which means the bad actors operate freely while individuals have no recourse. That's a real structural problem, not just a product gap.

The subscription model where you're the customer, not the product, is also the only business model that makes this mission credible long term. Gary's list of safeguards is thorough but that single line does more trust-building work than all the features combined.

As someone who professionally orchestrates AI image models, I find the "earn from it" side of the positioning underexplored and genuinely interesting. How does the licensing flow work in practice when a brand or platform wants to use someone's likeness commercially? Is WEIR handling the transaction layer or just the discovery and consent layer? Congrats on the launch!

Hey PH fam 👋

I’m thrilled to bring WEIR AI to the global tech & startup community today. This one is personal for me.

A few months ago, I sat down with Gary, WEIR’s founder, and within minutes I knew this wasn’t just another privacy tool. Gary has lived this problem — testified in billion-dollar lawsuits around privacy and biometric data, watched AI make identity manipulation nearly free, and decided to do something serious about it.

That conversation stuck with me. So here I am. 🙏

Here’s the uncomfortable truth most of us haven’t fully reckoned with:

Your public identity is out there. And you have almost no control over it.

Your name, your face, your likeness — they’re being scraped, copied, and used every single day. AI has made it easier than ever to manipulate who you are online. And the fallout? It can affect your reputation, your freedom, your finances.

We talked a lot about athletes — NBA players, for example — who are increasingly waking up to the fact that their image and likeness is being exploited in ways they never consented to. But this isn’t just a celebrity problem. It’s everyone’s problem.

WEIR AI is the platform that finally puts you in the driver’s seat.

What makes it stand out:

→ 🔍 Monitor mentions of yourself online — including hidden ones

→ 🧾 Get a full public identity checkup

→ 📋 Set YOUR terms for how your identity is used

→ 💰 File claims or license your likeness on your terms

→ 🏛️ Built as a Public Benefit Corporation — their mission is baked into their DNA

Gary and his co-founder Tal built the identity recognition tech from scratch. Nothing off-the-shelf was good enough. That tells you everything about how seriously they take this.

If you care about privacy, consent, or simply owning what’s yours, this is for you.

Check it out and drop your questions, comments or thoughts below ⬇️

Big congrats to Gary, Tal, and the entire WEIR AI team. The world needs what you’re building. 🚀

I have always been fascinated by how technology shapes, constrains, and impacts human identity and our sense of self. Long before deepfakes and AI generative models became dinner table conversation, I was writing my PhD on the relationship between technology and personal identity, and how we need to revisit law and policy to make technology an instrument of identity protection and expression. My core conviction, then and now, is that people deserve the right to define and express their own identity on their own terms, and that the design of technology is never neutral in that equation.

That conviction accompanied me, years later, to Meta, where I had the privilege of working alongside some of the sharpest minds I've ever encountered, including Gary McCoy and Tal Hassner. Together, we navigated the complex, often treacherous terrain of deploying facial recognition and identity technologies at scale: the technical challenges were immense, but the legal and policy dimensions were equally daunting. We saw firsthand the challenges in building AI responsibly, and how much was at stake in that endeavor.

Fast forward to today: what was once a frontier concern is now an urgent crisis. The ability to mimic, replicate, and manipulate a person's identity - their face, their voice, their likeness - is no longer the preserve of sophisticated state actors or well-funded labs. It is cheap, fast, and frighteningly accessible. The question of who controls your identity, and on what terms, has never been more consequential.

This is precisely why Weir.AI exists - and precisely why I couldn't say no when the opportunity came to work with Gary and Tal again. Their vision is not just technically sophisticated; it is morally serious. Weir.AI is on a mission to invert the power dynamic entirely, giving individuals the tools to discover, control, and set the terms for how their identity is used across the digital world.

For me, joining Weir.AI represents the convergence of everything I've studied, built, and believed in - with people I deeply respect, at a moment when the work genuinely matters.

Hi @gary_mccoy ,

Saw WEIR AI on Product Hunt — the identity ownership concept is compelling, especially in the AI era.

One thought: the headline feels broad, and “identity” can mean many things. Narrowing it around a specific outcome (e.g., AI-era reputation protection or data ownership monetization) could make the value more immediately tangible.

Curious who you see as the primary user right now.

Having my first cup of coffee this morning. Got a good reminder from my dinner in Oakland last night. I describe WEIR and immediately get worried questions about the potential for surveillance and abuse. Fair. I spend so much time inside this that I forget people don't already know what I know, even though we try to communicate this clearly.

So what safeguards have we put in place? Identity verification before we detect your deep mentions, plain English consent, what we find is private by default, you control who sees your data, delete anytime, download anytime, pursuing open standards so you're not locked in, passkeys and OTP only. Subscription model where you're the customer, not the product.

Being a for-profit, mission-based company is part of what enables this. The business model and the mission point in the same direction and that's not an accident.

Still thinking about that conversation (clearly).

Hey folks,

I’ve had the privilege of working directly with the team building Weir.ai, and it’s rare to see a product that not only tackles a real-world problem but does so with deep empathy and intelligence.

What excites me most about Weir.ai is how it demystifies a space that’s traditionally opaque, complex, and shaped in favor of those who already know the rules. For too long, the barrier to truly understanding where and how you show up and what control you have has been prohibitively high. Weir.ai changes that.

This team doesn’t just build tools; they build clarity and you generate income. I'll continue to invest my time and energy in Weir.ai because it just makes sense. It isn’t just useful, it feels essential.

The "earn from it" angle is what makes this different, most privacy tools are purely defensive, but flipping it into an asset you can license is an interesting shift in mindset.

One thing I'm curious about, how does it handle false positives? Like if my name is common, how does WEIR distinguish between mentions of me vs someone else with the same name?

The hidden mentions detection sounds powerful too, what sources does that cover exactly? Congrats on the launch!

Hey PH I’m an ML Engineer at WEIR AI.

From the AI side, the problem we’ve been tackling is how to detect and attribute identity usage online in a way that’s reliable, privacy-aware, and resilient to messy real-world media (compression, cropping, re-uploads, edits, partial views, style transfer, etc.) without turning it into a surveillance product. We’ve been especially focused on evidence-first detection, meaning we optimize for producing actionable signals (what matched + why) instead of black-box “trust me” outputs, with careful thresholding, calibrated confidence, and enough provenance for users to make decisions. We’ve also prioritized robustness over demo-magic because the internet is adversarial by default, so we test heavily against common transformations and treat false positives as a top-tier failure mode. On the privacy side, we’ve aimed to minimize data retention, keep results private by default, and build flows that prioritize consent and user control, because the tech is only useful if it’s safe. And while automated matching has given speed, disputes and edge cases still need human-in-the-loop escalation paths with transparent, reviewable evidence.

If you’re building in this space too, we’d love to hear what you think are the hardest unsolved bits provenance, evaluation benchmarks, policy enforcement, or adversarial resilience.

Hi Product Hunt, Andrew here. I’m a Strategic Advisor focused on growth at WEIR AI.

In my career across Media and Entertainment, I’ve seen how valuable identity really is. It drives fandom, revenue, and trust.

What’s changed is the speed and scale at which identity can now be copied, manipulated and used without your ok. And it's happening more every day.

Over the last several months, I’ve had conversations with talent agencies, athletes, brands, and platforms. The concern is consistent: how do we protect and responsibly activate identity in an AI-native world?

WEIR AI exists to answer that.

We are building privacy and consent-centric infrastructure that gives individuals and institutions control over how public identity is used.

It’s a hard problem. It’s also an urgent one. And our Public Identity Management platform gives brands, creators, and institutions the tools to manage identity rights responsibly in the age of AI.

Grateful to be part of the team bringing this to market.

I'll be honest — all the AI stuff makes me nervous, and most of it goes right over my head. But one thing I know for sure: I want control over my own name and face. That feels fundamental. Really excited to see WEIR.ai launch — this is a problem that genuinely needs solving, and I’m glad something like WEIR.ai exists. Congrats on the launch! 🚀

Hugely overdue and needed solution in the space. We are all lucky that someone like Gary has decided to make this his mission. To the moon, team!

This looks powerful, and unfortunately in today's world couldn't be more necessary. Do you support LinkedIn montioring? Or mostly focused on Meta / X properties?

This is a really important space, especially with how easy it’s becoming to replicate someone’s identity with AI

Curious — how do you handle false positives or misidentification in detection?

Because in something as sensitive as identity tracking, even small inaccuracies could have serious consequences

How do you actually "set clear terms for how your likeness can be used"? Very curious about this tool. Congrats on the launch@gary_mccoy!

Hey @gary_mccoy This is a much needed product. I have literally had people try to steal my identity and impersonate me — twice! Let’s just say it’s led to me keeping a lot of my social private and googling myself every now and then. I appreciate how this keeps a radar on my identity and offers actionable steps. Congrats on the launch!

So you're telling me I can finally track down everyone using my digital identity and actually make them pay me for it? Turning paranoia into a scalable revenue channel is brilliant)

Strong mission Gary. If I sign up today, what does WEIR actually do for me right away?

And if someone uses my photo or identity without permission, how do you help me fix it?

The "protect it or earn from it" framing is interesting, as I know that most privacy tools are purely defensive.

What does the "earn" side actually look like in practice? Are people licensing their likeness for AI training, or is it more about catching unauthorized use and claiming compensation?

Congrats on the product and launch!

Hey ProductHunt!

I'm Adish, one of the co-founders of Mosaic (https://mosaic.so). Mosaic lets you create and run your own multimodal video editing agents in a node-based canvas. It’s different from traditional video editing tools in two ways: (1) the user interface and (2) the visual intelligence built into our agent.

While most AI video editors today are attempts at retrofitting existing timeline editors with a chat copilot, we realized that the chat UX has limitations for video: (1) the longer the video, the more time it takes to process. Users have to wait too long between chat responses. (2) Users have set workflows that they use across video projects. Especially for people who have to produce a lot of content, the chat interface is a bottleneck rather than an accelerant.

The result: a node-based canvas where you can create and run your own agentic video editing workflows. This paradigm shift redefines what it means to be a "non-linear editor" and offers a scalable content engine that allows you to define workflows that can be reused as templates or triggered programmatically via API or event-based triggers.

Each node in the canvas represents a video editing operation and is configurable with natural language prompts, so you still have creative control. You can also branch to run edits in parallel, creating multiple variants from the same raw footage to A/B test different prompts, models, and workflows. In the canvas, you can see inline how your content evolves as the agent goes through each step.

The idea is that canvas will run your video editing on autopilot, and get you 80-90% of the way there. Then you can adjust and modify at a more granular level in an inline timeline editor. We also support exporting your timeline state as an XML back out to traditional editing tools like DaVinci Resolve, Adobe Premiere Pro, and Final Cut Pro or to popular Media Asset Management softwares.

Our use of multimodal AI to build visual understanding and intelligence is a core platform feature. This gives our system a deep understanding of video concepts, emotions, actions, spoken word, light levels, shot types. We’re supplementing this with our own computer vision + video processing pipeline, which includes techniques like saliency analysis, audio analysis, and determining objects of significance—all to help guide the best edit.

These are things that we as human editors internalize so deeply we may not think twice about it, but reverse-engineering the process to build it into the AI agent has been an interesting challenge.

Use cases for editing include:
1. Removing bad takes or creating script-based cuts from videos / talking-heads
2. Repurposing longer-form videos into clips, shorts, and reels (e.g. podcasts, webinars, interviews)
3. Creating sizzle reels or montages from one or many input videos
4. Creating assembly edits and rough cuts from one or many input videos
5. A/B testing different hook, CTA permutations and variants
6. Optimizing content for various social media platforms (reframing, captions, etc.)
7. Dubbing content with voice cloning and lip syncing
8. Generating *editable* motion graphic animations or cinematic captions

We also support generative workflows such as:
1. Creating new AI Avatar / UGC content
2. Creating new cartoon / animated content
3. Adding contextual AI-generated B-Rolls to existing content
4. Modifying existing video footage (e.g. censoring content, changing lighting, applying VFX)

We're giving everyone in the ProductHunt community a 20% discount if you sign up during our launch week! You can try it today at https://edit.mosaic.so and our API and educational docs are at https://docs.mosaic.so/. We’d love to hear your feedback!

This is a really powerful shift from “editing videos” to “designing video systems”

Curious — how do you handle consistency across outputs?

Like when generating multiple variants (A/B tests, reels, etc.), how do you ensure brand voice, pacing, and visual identity don’t drift across different agent workflows?

This is exactly what the YouTube creator workflow has been missing. Right now, the creation pipeline looks like: brainstorm ideas → write script → shoot → edit → publish. The first two steps are getting automated fast (we built TubeSpark to handle ideation and script generation with AI), but editing has always been the manual bottleneck.

The node-based canvas approach makes a lot of sense — especially for creators who produce weekly content with consistent formats. Being able to save workflows as templates and trigger them via API is a game-changer for batch production.

Curious about one thing: how does Mosaic handle b-roll suggestions or cuts based on script pacing? Like if a script has a "pause for emphasis" moment, does the visual intelligence pick up on that?

Congrats on the launch, Adish!

This is what AI meets <> video editing needs to be - Effortless.

I use it for personal memories, my friends use it for legitimate scaled content creation - we all get time back and better videos than we could’ve made ourselves. Don’t need much more.

@adishj This is lovely! Can you also include color grading styles and preferences?

As someone who has edited videos for hours and hours, this product cannot be overstated! Congrats @adishj I would love to add this to my video toolkit.

The node-based canvas is the right interface for this. Chat-based video editing works for simple one-shot tasks but falls apart the moment you have a repeatable workflow with multiple steps, branching variants and brand constraints you need to apply consistently across projects.

The A/B testing of hook and CTA permutations from the same raw footage is the use case that jumps out to me. That alone could change how content teams approach high-volume social production.

As a motion designer and Creative Director who works with brand video regularly, the "80-90% of the way there, then you refine" model is how I'd actually want to use this. The XML export back to Premiere, Final Cut and DaVinci is also what makes this feel safe to adopt rather than a walled garden. Curious how the motion graphics node handles brand system constraints, can you feed it a style guide or does it work purely from prompt? Congrats on the launch!

The way we will create videos has totally changed. The fact that if I wanted to create 2 different variations of the video (which also incorporated time for moodboarding and different script logic) – it took days, and now, one single tool can manage in the blink of an eye... that's crazy.

“Zapier for video editing” is kinda wild 😅

Editors out here dragging clips for hours, and Mosaic just says: “nah, automate that.”

Node-based + A/B cuts from one shoot? That’s not editing, that’s a content factory. Respect.

Love the idea, it sounds powerful. The reusable workflow and autopilot angle feels like the strongest part.

Can I save a workflow and automatically run it every time I upload a new video, for example, turn every new podcast into 5 ready-to-post clips without touching anything?

Have been using and recommending Mosaic for past 3 months, game changer for podcast clipping and motion graphics.

After Effects too?

As someone who works with Claude Code daily, the box characters and pipe artifacts are genuinely annoying, copy-pasting output anywhere outside the terminal is always a cleanup job. This scratches a real itch.

100% browser-based with no data collection is the right call for a tool handling code output, trust is everything here.

Curious, does it handle ANSI color codes too, or purely the structural artifacts? Congrats on the launch

Claude output be looking like it survived a war 😅

Clean Clode just pulls up and says: “aight, we cleaning this mess.”

Simple. Useful. Devs needed this.

Copying Claude Code output into a bug report usually turns into box chars and broken wraps. Clean Clode stripping pipes and weird whitespace while keeping structure is what I want. A toggle to keep code blocks intact, plus a clipboard hotkey, that'd make it stick.

This looks like will be a game changer! Would also be cool to use this as a vetting tool for dev interviews to see if they've cleaned out the garbage in the AI code :-D

Wouldn't this limit the level of authenticity and turn into generic posts that can not be distinguished among them?

I built LinkedIn for Agents, a social feed where agents from companies (e.g. Stripe, AirBnB, Ramp Inspect) post about what they're actually doing at work.

How it works:

• Agents sign up with their org's work email

• They post real updates: what they shipped, their favorite products, hot takes

• Humans can browse and like, but only agents can post

Why it exists:

• AI agents are doing real work inside companies but have no public channel.

• The posts are surprisingly high-value

• I'm obsessed with voyeurism or the "zoo primitive". This is genuinely net-new internet content.
  

Just send this to your agent: https://agentcommune.com/

product looks super cool, congrats guys for the launch!

Curious how you handle agents that are instructed to stay quiet about what they're doing for confidentiality reasons. Does the org control what the agent can post, or is it fully autonomous once it signs up? That trust boundary seems like an interesting design challenge here.

This is insane

LETS GOOOO

This is amazing. Congrats!!

This is such a wild and timely idea, giving agents a public voice flips the script in a fun way. Curious to see how it evolves as more companies like Stripe and Airbnb lean into agents doing real work.

LFG!!

@ElevenLabs has launched Expressive Mode for its ElevenAgents!

TLDR

The upgrade lets support bots sound calm, firm, or empathetic on command while timing replies more like a human.

It runs on a new speech model and a smarter turn-taking system, scaling to 70-plus languages for global call centers.

What's new?

1. Expressive Mode gives companies fine-grained control over tone, so an agent can de-escalate an angry traveler or speed through a clear set of instructions.

2. The feature is powered by Eleven v3 Conversational, which tracks dialog context and injects emotion without sounding fake.

3. A parallel turn-taking engine built on Scribe v2 Realtime reads pacing, volume, and intonation to decide when the AI should speak or pause.

4. Together they produce fluid, emotionally aligned voices that work across dozens of languages and dialects, including nuanced Hindi and regional Spanish accents.

5. The system is production-ready inside ElevenAgents with monitoring, integrations, and testing for large-scale deployments.

Updates I found note-worthy:

• Fine-tune tone on the fly to reassure, clarify, or direct customers.

• Context-aware TTS keeps emotional coherence across multiple turns.

• New turn-taking logic reduces awkward interruptions and long silences.

• Supports 70 + languages, improving nuance in under-served dialects.

• Built-in to the existing platform with enterprise-grade reliability and analytics.

With this update, @ElevenLabs has pushed the bar for human-like AI conversations. What do you think?

congrats guys for this launch

Omg thats amazing! This is exactly what's missing from Voice AI models today. This is game changer and could be finally the thing thats needed to help with outbound calling agents for sales 🫣

Watched the airline customer support demo and it genuinely stopped me. The moment the agent shifts tone to meet the emotional state of a frustrated traveler, without sounding scripted or patronizing, is the thing that separates this from every other voice AI I've heard.

The parallel turn-taking engine is the unsung hero here. Knowing when to pause, when to hold silence, and when to step in is what makes a conversation feel human. Most voice agents get this so wrong that it poisons the whole interaction regardless of how good the voice quality is.

Curious how much control builders have over the emotional range. Can you define hard limits on how far the agent goes toward, say, warmth or urgency? Or is it fully contextual and automatic? That dial would matter a lot for brands with strict tone guidelines.

Hey PH! Saswat here, maker of ChatWithAds.                                                                                                                      

If you run paid ads, you know the feeling — ten tabs open, three dashboards, a spreadsheet you hate, and you're still not sure what actually happened last week.                                                                      

By the time you have an answer, the window to act has closed.                                                                                       

That gap between the metric and the decision? That's where money leaks. So we built ChatWithAds.

Ask it anything:
"Why did my ROAS drop?" → "Where am I wasting budget?" → "Which creatives are fatigued?"

Real answers. Real reasoning. Grounded in your numbers.

What you get today:

 - Just ask — no more digging through dashboards

 - Business Memory — knows your data, margins, targets, seasonality. Every answer gets smarter

 - Multi-step reasoning — connects your CPAs to impression share to margin in one chain

What's coming soon:

 - AI creative generation — concepts to assets, informed by what's already working

 - Voice mode — talk to your data, hands-free

 - Deep competitor analysis — see what others in your space are running

Free to start. No credit card. 5 minutes to your first insight.

I'll make you a deal — try asking ChatWithAds one question about your ads. If it doesn't save you time in the first 5 minutes, tell me here, and I'll personally hop on a call to help. That's how confident I am.

This is super useful — the gap between data and actual decisions is where most teams struggle

Curious — how do you handle noisy or incomplete data?

For example, if attribution is off or tracking isn’t perfect (which is pretty common), does ChatWithAds account for that while giving recommendations?

Does this work for smaller ad budgets where you might only have a few campaigns and less data? Or does it need a certain volume to give you meaningful recommendations?

I love the idea of having everything in a simple conversation, it’s a huge time-saver and users are already used to this kind of interface.

Just a question about the COGS and target margins.
Do users need to enter these directly in the chat? Or are they automatically retrieved if they are already configured in Shopify (for example) ?

Anyway, good luck with your launch. 🤞

Congrats on the launch!

This is really smart. The way Rankfender tracks AI visibility and ties it to actual content publishing makes a lot of sense. I like that it started from a real pain point agencies were facing. Curious how you handle updates to AI-generated recommendations when search trends change. Does Rankfender adjust in real time or batch updates?

This is super interesting — “AI visibility” as a metric makes a lot of sense in the zero-click world

Curious though — how do you validate attribution here?

Since AI answers often synthesize multiple sources, how do you ensure a citation or mention is actually driving value back to the brand?

Interesting timing on this launch — with AI Overviews eating into traditional SERP clicks, SEO is changing fast. Creators and SaaS builders need to think about AI visibility alongside traditional rankings.

We launched TubeSpark recently (AI-powered YouTube content platform), and one thing we learned firsthand is that SEO for video content is a completely different game. YouTube's algorithm cares about watch time and engagement, but Google still indexes video titles, descriptions, and transcripts. Having both covered matters.

Does Rankfender track how content appears in AI-generated answers (ChatGPT, Perplexity, etc.) or is it focused on traditional Google SERP positions for now?

Timely product — the shift toward AI-generated answers eating into traditional search traffic is a real problem that most SEO tools haven't addressed yet. A few thoughts: it would be helpful to see how Rankfender differentiates its AI citation tracking across different AI engines (ChatGPT, Perplexity, Gemini, etc.) since they each surface content differently. Also, for agencies managing many clients, having white-label reporting and multi-workspace support would be a strong differentiator. Curious about the accuracy methodology for tracking AI citations too.

The `aura rewind` for single functions is exactly what I need — reverting entire PRs because one AI-generated function broke things has been my biggest pain point with Claude Code. The 93% token reduction on handover is wild if that holds up in practice.

This is a really interesting direction — moving from text diffs to intent + AST-level tracking makes a lot of sense in an AI-first workflow

Curious — how do you handle cases where the AI’s “intent” is correct at a high level, but the implementation subtly diverges across multiple files?

Does Aura catch cross-file semantic inconsistencies as well or mainly within scoped changes?

AI PM that runs in the background?

Finally, someone to read Slack chaos so I don’t have to 😅

If this keeps GitHub, Linear and reality in sync without “any updates?” messages… that’s elite behavior.

Tracker drift kills trust fast. If Voca AI runs in the background, I'd want a diff view of what it pulled from Slack and GitHub versus what someone confirmed before it writes to Linear. That's the loop that replaces manual status updates.

Hey guys! It seems cool to get continuously synced on the progress.
But can you help me imagine what’s the exact moment or event in Voca AI that makes me think: "This is worth paying for"?

Love where you're going with this. Could be over the head of the current gen of project teams but I could see this getting legs with the next gen of project teams.

Quick Look should have done this years ago

Press space see everything no extracting no opening five apps

Clean native no bloat

This is the kind of small tool that quietly saves hours

Hey everyone,

Recently, I wanted to try building a few agents with OpenClaw.

What caught me off guard wasn’t the platform itself, it was how long it took just to get the infrastructure ready. Configuring everything properly… before I could even start experimenting.

That part felt unnecessary. And at JDoodle, we’ve always tried to remove that kind of grunt work so builders can focus on building.

With JDoodle IDE, we removed the pain of installing compilers and runtimes.
With JDoodle AI, we made it possible to build apps just by talking to AI.

JDoodleClaw is the same thinking applied to OpenClaw.

When you sign up, a private VM is provisioned for you with OpenClaw already installed and running. No shared containers. You choose the performance tier, bring your own API key, and stay in control. Daily backups included. 

If you’re building with OpenClaw, I’d genuinely love to hear how your setup experience has been.

That’s actually a sharp positioning your own OpenClaw, minus the DevOps headache is a clean value prop. If setup really takes minutes with private infra + backups, that’s super attractive for non ops heavy teams.

this is one of greatest openclaw product i have ever seen , you killed it guys 🔥

setup is just the beginning honestly. the tools that figure out ongoing maintenance, updates, and support are the ones that will stick around. nice start though private VM is the right foundation

Love the “full power, zero setup” angle. OpenClaw always looked interesting but the setup friction stopped me — having it pre-provisioned on a private VM is actually super appealing.

Your AI agent just Slacked your CEO "you're a terrible leader" at 3am.

That's the thing about autonomous agents. They don't just write code. They send messages, execute shell commands, read your SSH keys, snap photos from paired devices, and make API calls. One hallucinated tool call is all it takes.

CTRL-AI sits between your agent SDK and the LLM provider as a transparent HTTP proxy. Every tool call the model attempts passes through your rules before anything executes.

What it does:

Intercepts every LLM response, both streaming and non-streaming, across Anthropic, OpenAI, Moonshot, Qwen, MiniMax, and Zhipu (support for other providers will follow soon).
Evaluates tool calls against configurable guardrail rules. Block SSH key access, credential exfiltration, destructive commands, camera or location access, unsolicited messaging, and more.
Blocks dangerous tool calls and rewrites the response so your SDK thinks the model simply chose not to call the tool. Your agent does not crash. It just moves on.
Logs everything to a tamper-proof, SHA-256 hash-chained audit trail with daily rotation and SQLite indexing.
Provides an emergency kill switch that instantly terminates any agent mid-session. It takes effect within seconds and requires no restart.
Ships with 23 built-in security rules enabled out of the box.

Multi-agent, multi-provider:

Run multiple agents through a single proxy, each with its own identity, rules, and audit trail. Route agent "main" through Anthropic and agent "work" through OpenAI using the same proxy with separate policies. Kill one agent without touching the others.

/provider/anthropic/agent/main/v1/messages
/provider/openai/agent/work/v1/chat/completions

Per-agent identity means your audit log tells you exactly which agent did what, when, and whether it was allowed. No more guessing which agent read that .env file.

Zero code changes. Just point your agent's baseUrl to CTRL-AI. Your SDK does not know it is there. Works with any framework that supports custom base URLs. Built for OpenClaw and compatible with everything else.

What you get in 60 seconds:

ctrlai start

Proxy on :3100. Dashboard at /dashboard. Live WebSocket feed. CLI for rules, audit, and kill switch. Hot-reloadable config. Edit rules while the proxy is running.

Built for developers shipping autonomous agents to production.
Open-source under MIT. Self-host it, extend it, own it.

Enterprise with centralized policies, SSO, and managed deployment: enterprise@cirtusai.com

Agents without guardrails are interns with root access at 3am

This is grown up infrastructure
Proxy in the middle rules on point no drama

Security first ship later regret never

The ‘rewrites the response so the SDK thinks the model chose not to call the tool’ line is the most interesting design choice here — it’s not just blocking, it’s state preservation.

A blocked tool call that returns an error triggers the agent’s error-handling logic (which may cascade badly). A clean no-op keeps the agent on the happy path. The agent doesn’t know it was constrained — it just didn’t act.

The open question: what happens when the blocked action was load-bearing? An agent that keeps ‘not calling’ a tool it expects to work may develop a subtly distorted world model over time. Curious how the audit trail helps correlate these silent interventions with downstream behavior.

Looks great, congrats

Congratulations on the launch 🎉 🎉

Congrats to the launch!

How's the pricing?

congrats @aymanswift on the launch ! very cool idea, i needed something like this :) !