Jung von Matt CREATORS

Imagine – it's Fashion Week and I talk about, how to dress 600,000 people in sweatpants? Panel talk with Anna Franziska Michel from yoona.ai (Techstars ´21) and Fax Quintus from i-mmersive. #GenAI #JvMSTABLES #sweatpantsday 🎤 yoonaverse 4.30 pm 📍 Tucholskystr. 2, Google Berlin

Last week, JvM CREATORS took part in the HORIZONT Prompt Battle. The brief: What would a DOOH campaign for the brand Germany using the help of AI tools look like? While we really liked our idea of #allealman for the brand Germany, we didn’t win – even though we used Midjourney, Luma, Runway, Magnific, and even Cinema4D. However, not all is lost. We’re not sore losers. JvM STABLES PRESENTS: LESSON LAIRNED! Let’s talk about 5 lessons we learned during these 2.5 hours of intense AI usage. ⬇️ Do you agree, disagree or did we miss anything? Drop it in the comments! 🎥👁️ Big thanks to everyone involved: Michael Schlykow (häppy) Bastian Goldschmidt (Grabarz & Partner) Robert Andersen (Jung von Matt CREATORS) and congrats to Paulina for winning the Prompt Battle! 🥊 #genAI #JvM #lessonlairned #Runway #Pika #HORIZONT #promptbattle #ai #midjourney #luma #magnific #cinema4d #allealman Jung von Matt CREATORS

My prompt battle learning: Don't attempt to rebrand Germany in just 2.5 hours, even with #GenAI. But if you did, well done: Paulina, Bastian, and Michael. 🙏 Thanks to all who voted for #ALLEALMAN

Let's check out some talking heads. Not the band, you beautiful Gen-Xers. 🤫 AI-animated characters are becoming more and more popular with new tools popping up all around. So we did the only logical thing: Put four of them to the test by making them reenact some of Germany's most legendary #memes. Can they beat the originals? Of course not, but check out the video to see which model came closest! ⬇️ JvM STABLES PRESENTS: LESSON LAIRNED! 💃 HEDRA Hedra's lip sync is lively and funny to watch. Nevertheless, more control over expressiveness is needed, as the eye movement is excessive and sometimes odd. Additionally, the output needs to be larger than 512x512. It's a promising beginning, and we look forward to future updates! 💡 RUNWAY Runway, on the other hand, is much more static. It makes fewer mistakes and unwanted movements, but it often feels like you're talking to a puppet. Furthermore, Runway adapts dynamically to the image size. The tool is more functional because it makes fewer mistakes than Hedra, and with more expressiveness, it would be great. ♠️ D-ID D-ID has been around for a while and offers a good compromise between Hedra's and Runway's performance. The faces get animated well with some control of the facial expression, while the lip syncing also does a decent job. It does however lack the liveliness of Hedra's best generations. It's a very solid choice overall. 🧚🏽♀️ MUSEV + MUSETALK MuseV + MuseTalk have the great advantage that they are open source and can be used for free - if you know how to set up the software locally. That does require a bit of tech-savviness. But once you're set-up, it's relatively easy to use and offers prompt control of the animation. Unfortunately, it does not perform well across the board. Realistic images perform much better than paintings or drawings and the lipsyncing does not operate at a very high resolution, which shows. But it's a valuable contribution nonetheless and with some tinkering can be a good and free alternative to commercial products. There are some more open-source alternatives, at which we might look again in the future. Check out our comparison and see for yourself! 🎥👁️ #genAI #JvM #lessonlairned #Runway #Hedra #DID #MUSEV #MUSETALK

Attending #K52024 in Berlin? Let's connect and discuss #GenAI: embracing new mindsets and rethinking customer experiences 🤞. See you at the #DeepDive Stage with industry leaders: Johannes Altmann, Erik Reintjes, Joscha Krug 🎩, Laura Lewandowski 🎙, Tim Böker 🙌 Tuesday, June 25th 👁 2:30 pm, Estrel Berlin P.S. Can we meet Mirja Meyer, Thomas Feldhaus & Jana Stieler?

Wake up, babe! A new Runway model just dropped – actually, it’s still not public yet! ❤️ Anyway, just because we are all yearning for a reliable video GenAI with object persistence and 4K resolution, it doesn’t mean that its newest release will live up to its promises (hinting at you, Stable Diffusion 3 😉). That’s why we’re comparing it to two competitors: Luma AI and Pika. Roll it! JvM STABLES PRESENTS: LESSON LAIRNED! 💃 RUNWAY These Runway Gen 3 videos are impressive but cherry-picked. With Gen-3 not publicly available yet, independent testing isn’t possible. However, the AI shows good coherence for up to 5 seconds and achieves realistic rendering. One step closer to a production-ready solution! 💡 LUMA AI LumaLabs' new video AI model expands images with shaky coherence and questionable rendering. Text-to-video prompting is even worse, often hallucinating with creative prompts. It's more a glimpse into the future than a production-ready solution! ♠️ PIKA Pika excels at image coherence with minimal motion, acting more as a zoom-in tool than a storytelling device. Videos from provided images retain quality, but those from prompts struggle with realistic humans. It's more a niche gimmick than a practical production tool. Check out our comparison and see for yourself! 🎥👁️ #genAI #JvM #lessonlairned #Runway #LumaAI #PikaLabs #Pika

Wake up, babe! A new Runway model just dropped – actually, it’s still not public yet! ❤️ Anyway, just because we are all yearning for a reliable video GenAI with object persistence and 4K resolution, it doesn’t mean that its newest release will live up to its promises (hinting at you, Stable Diffusion 3 😉). That’s why we’re comparing it to two competitors: Luma AI and Pika. Roll it! JvM STABLES PRESENTS: LESSON LAIRNED! 💃 RUNWAY These Runway Gen 3 videos are impressive but cherry-picked. With Gen-3 not publicly available yet, independent testing isn’t possible. However, the AI shows good coherence for up to 5 seconds and achieves realistic rendering. One step closer to a production-ready solution! 💡 LUMA AI LumaLabs' new video AI model expands images with shaky coherence and questionable rendering. Text-to-video prompting is even worse, often hallucinating with creative prompts. It's more a glimpse into the future than a production-ready solution! ♠️ PIKA Pika excels at image coherence with minimal motion, acting more as a zoom-in tool than a storytelling device. Videos from provided images retain quality, but those from prompts struggle with realistic humans. It's more a niche gimmick than a practical production tool. Check out our comparison and see for yourself! 🎥👁️ #genAI #JvM #lessonlairned #Runway #LumaAI #PikaLabs #Pika