Experience storytelling like never before. ScaryStories Live transforms your ideas into live, AI-generated story scenes that evolve as you direct them. Perfect for creators, streamers, or anyone exploring the future of interactive storytelling.
Unblocked AI Code Review — High-signal comments based on your team's context
High-signal comments based on your team's context
Promoted
Maker
📌
Hey folks,
I’m Ning, ex-FAIR researcher and Llama 3 Speech contributor. I was also a founding member of Seamless Communication (AI translation) which was featured on Nature and TIMES best innovations 2023.
My background is in real-time multimodal systems: making models not just see or speak, but respond. I'd like to build a narrating layer for today's AI videos.
I love the AI Dungeon experience because it gives you true agency: the story is yours. And recent wave of AI models unlock new experience. But wait, latency kills the mood. If the monster waits 3 seconds to move, it’s comedy. And hopping between prompts like stitched stills break immersive experience. These are the two problems we’re fixing:
- Latency: we aim for ≤3 seconds p95, meaning 95% of reactions land in under three seconds to keep the rhythm tight. We’re also building full streaming control so you can steer the scene with arbitrary speech or text prompts as it unfolds.
- Camera movement: instead of hopping between prompts like stitched stills, we turn choices into shots: reliable pans, tilts, dollies, rack focus, and smooth eases, while preserving screen-space anchors and temporal continuity.
For me, this project is a way to turn all that research into something visceral where performance and machine cognition blur together in real time.
@ning_dongHow are you tackling the model overhead and maintaining visual continuity shot to shot?
Report
Maker
@mohsinproduct Great question! If model overhead means latency here, we adopt pixel->token compression rate that's on the extreme side. While it's hard to build a model for general purpose, with careful lightweight finetuning for specific niche (like horror as our first probe) it works reasonably well. For visual continuity it is also a hard problem, many research work trying to tackle this recently and optimization could also be added in production layer.
I'm Helen. I helped ship AI games to millions of players and built creative tools for more than 300K professional creators.
ScaryStories Live is AI Dungeon meets real-time video generation. If AI Dungeon lets you play through your imagination with text, ScaryStories Live lets you see it, hear it, and feel it, in real time. You direct the AI, make choices, type your next move, and within milliseconds, the scene comes to life with sounds, whispering voices, and visual clues. 😈
We built ScaryStories Live because no matter how many AI tools we have today that generate “beautiful” videos, most of them are just that beautiful, not alive. They don’t make your heart race, your hands sweat, or your imagination run wild.🥱 Great stories come from the process of discovery, not from static prompts or templates. So we designed ScaryStories Live to make creation fun, fast, and alive.😊
Imagine watching Stranger Things and suddenly stepping into the Upside Down yourself, shaping the outcome, recording your journey, and sharing it as a cinematic short that feels alive, emotional, and terrifying. Or uploading a photo from your bedroom, then experiencing what happens next, in real time. Once your adventure ends, our AI helps polish it into a high-quality short video, with pacing, sound, and emotion tuned perfectly, ready to share with the world and our community.🙌
We’re starting with scary stories, but our vision is bigger: to turn ScaryStories Live into a new medium where any genre, any fandom, and any creator can bring their imagination to life, not by writing prompts, but by living them. 💕
@sanskarix Thanks so much, Sanskar! We're so stoked you think so. We couldn't resist a Halloween drop for those perfect spooky vibes! 😉 Thrilled you're digging the immersive feel. That's exactly what we were going for. Happy Halloween!👻
What a day to launch scarystories!! i'm so happy to see this project come to life and fish.audio is honored to power such a well made app. congrats team!!
@hehe6z Thanks so much, Helena! ❤️ That really means a lot coming from you and the Fish Audio team. We're beyond stoked to have you powering our stories. So glad we could get this out on the spookiest day of the year! 🎃 Happy Halloween! 👻
Report
Maker
@hehe6z Thanks Helena! I've been working in the speech field for a few years and some voices in fish.audio are the most expressive ones I heard, which fit our scenarios the best!
@jaredl Thanks so much for the support, Jared! Right now, you can download the final, polished MP4 to share anywhere (like TikTok or YouTube), or you can publish it directly to our community for others to play and remix. Our AI engine is about handling all the "hassle parts" for you. 🙌 Think of it as AI Dungeon meets real-time video generation, where we deliver a finished gamified result directly to you. While a raw timeline export isn't available, we're massively expanding creative control. In upcoming versions, you'll be able to guide the final output by customizing the AI character's image/voice, setting goals, adjusting memory, selecting models, and more! 🤩
@mona_xx Thanks so much! Launching on Halloween felt perfect. Like a ghost storyteller, but with real-time chills, exactly the vibe we’re chasing. Appreciate the love!
@cruise_chen Thank you so much, Cruise! We're thrilled you think it's cool! 😊 That's the perfect question! So, right now, after you direct the AI through the 5 interactive sketch steps, the final, auto-magically polished MP4 (with the voiceover, sound effects, pacing, etc.) is up to 20 seconds. BUT... that's just for now! We are incredibly excited to share that we're about to remove those limits. Very soon, we'll be rolling out a new version where you can blow past the 5-step limit and create videos that are as long as you want! We're talking a "freely exploring" version where your creativity is the only limit. We can't wait for you to get your hands on it!🙌
PicWish
@ning_dong How are you tackling the model overhead and maintaining visual continuity shot to shot?
@mohsinproduct Great question! If model overhead means latency here, we adopt pixel->token compression rate that's on the extreme side. While it's hard to build a model for general purpose, with careful lightweight finetuning for specific niche (like horror as our first probe) it works reasonably well. For visual continuity it is also a hard problem, many research work trying to tackle this recently and optimization could also be added in production layer.
ScaryStories Live
Hey Product Hunt Fam ❤️,
Happy Halloween 👻！
I'm Helen. I helped ship AI games to millions of players and built creative tools for more than 300K professional creators.
ScaryStories Live is AI Dungeon meets real-time video generation. If AI Dungeon lets you play through your imagination with text, ScaryStories Live lets you see it, hear it, and feel it, in real time. You direct the AI, make choices, type your next move, and within milliseconds, the scene comes to life with sounds, whispering voices, and visual clues.
😈
We built ScaryStories Live because no matter how many AI tools we have today that generate “beautiful” videos, most of them are just that beautiful, not alive. They don’t make your heart race, your hands sweat, or your imagination run wild.🥱 Great stories come from the process of discovery, not from static prompts or templates. So we designed ScaryStories Live to make creation fun, fast, and alive.😊
With ScaryStories Live, you’re inside your story.
🎮 You play it like a game.
🎥 You create it like a film.
💀 You feel it like a nightmare.
Imagine watching Stranger Things and suddenly stepping into the Upside Down yourself, shaping the outcome, recording your journey, and sharing it as a cinematic short that feels alive, emotional, and terrifying. Or uploading a photo from your bedroom, then experiencing what happens next, in real time. Once your adventure ends, our AI helps polish it into a high-quality short video, with pacing, sound, and emotion tuned perfectly, ready to share with the world and our community.🙌
We’re starting with scary stories, but our vision is bigger: to turn ScaryStories Live into a new medium where any genre, any fandom, and any creator can bring their imagination to life, not by writing prompts, but by living them. 💕
🤩 Try the best real time ai video generator for FREE Now！
We’d love your feedback to help us improve and grow. Let's shape the next frontier of storytelling together. 🚀
Best,
Helen Xiong
Fish Audio
@helen9 yayyy
Cal ID
What a perfect Halloween launch! This looks truly immersive.
ScaryStories Live
@sanskarix Thanks so much, Sanskar! We're so stoked you think so. We couldn't resist a Halloween drop for those perfect spooky vibes! 😉 Thrilled you're digging the immersive feel. That's exactly what we were going for. Happy Halloween!👻
@sanskarix Thanks Sanskar, happy Halloween!
ScaryStories Live
@sanskarix Thanks so much! Launching on Halloween felt right, thrilled the immersion lands. Appreciate the love! 👻
Fish Audio
What a day to launch scarystories!! i'm so happy to see this project come to life and fish.audio is honored to power such a well made app. congrats team!!
ScaryStories Live
@hehe6z Thanks so much, Helena! ❤️ That really means a lot coming from you and the Fish Audio team. We're beyond stoked to have you powering our stories. So glad we could get this out on the spookiest day of the year! 🎃 Happy Halloween! 👻
@hehe6z Thanks Helena! I've been working in the speech field for a few years and some voices in fish.audio are the most expressive ones I heard, which fit our scenarios the best!
YouMind
Congrats! Can creators export the raw timeline (camera parameters + audio) for further editing in external tools?
ScaryStories Live
@jaredl Thanks so much for the support, Jared! Right now, you can download the final, polished MP4 to share anywhere (like TikTok or YouTube), or you can publish it directly to our community for others to play and remix. Our AI engine is about handling all the "hassle parts" for you. 🙌 Think of it as AI Dungeon meets real-time video generation, where we deliver a finished gamified result directly to you. While a raw timeline export isn't available, we're massively expanding creative control. In upcoming versions, you'll be able to guide the final output by customizing the AI character's image/voice, setting goals, adjusting memory, selecting models, and more! 🤩
ScaryStories Live
X-Design
Just in time for Halloween! It’s like having a campfire ghost storyteller in your pocket, but way smarter and way more unsettling. Bravo to the team!
@mona_xx Thanks Mona, glad you like it!
ScaryStories Live
@mona_xx Thanks so much! Launching on Halloween felt perfect. Like a ghost storyteller, but with real-time chills, exactly the vibe we’re chasing. Appreciate the love!
Agnes AI
Super cool that ScaryStories provides an interactive storytelling! Just curious - what is the maximum length of the story does it support so far?
ScaryStories Live
@cruise_chen Thank you so much, Cruise! We're thrilled you think it's cool! 😊 That's the perfect question! So, right now, after you direct the AI through the 5 interactive sketch steps, the final, auto-magically polished MP4 (with the voiceover, sound effects, pacing, etc.) is up to 20 seconds. BUT... that's just for now! We are incredibly excited to share that we're about to remove those limits. Very soon, we'll be rolling out a new version where you can blow past the 5-step limit and create videos that are as long as you want! We're talking a "freely exploring" version where your creativity is the only limit. We can't wait for you to get your hands on it!🙌