AI Entrepreneurs’ Post

First try generations with Image to Video + same prompts in Runway Gen 2 (left) and Luma Dreammachine (right). I‘m wildly impressed by Luma AI Dreammachines result in the top right corner. It‘s a very complex scene to achieve and it did a great job to reconcile different factors: - stability of car interior - motion of point of view perspective - surrounding passing by - motion blur in general - exterior traffic - lead car driving same direction - oncoming traffic different direction - hands on wheel - rapid movements of steering wheel Even the needle of the speedometer is slightly moving, which is amazing. The scene of the car in rain turned out fine in both AIs. I personally prefer the result of Runway, because it added a slow dolly in. It suffered brilliance and sharpness for that, but it‘s more cinematic this way. I‘m very curious if Gen3 Image to Video will fill this gap and keep up with Dreammachine. Text to Video of Gen3 already looks promising in terms of movements and precision.

You use artificial intelligences for video generation? What are your preferences? Here's a simple comparison between Luma and Runway. #DreamMachine #LunaAI #runway #artificialintelligence

1 KEYFRAME to create 9 VIDEO CLIPS It all started this morning with the prompt "random image" and testing a combination of 2 newly found SREF STYLES. I was curious to see how Runway would generate different clip ideas into videos based of this sci-fi image. This video shows 9 VIDEO CLIPS generated in Runway Gen-3 Alpha. Some learnings from this AI VIDEO SESSION: - Runway can have problems with fingers ( most shots only 3 or 4 fingers ) - Runway tends to change the KEYFRAME image colors a bit - Runway impresses me in showing a correct "other side" when orbiting around objects / scenes - Runway's Unlimited plan may be the only "way" to explore video generation and generate as many clips you want to choose from the best results - ElevenLabs Webtool ( musictosoundeffects-com ) produces quite useable soundeffects for your scenes Both SREF STYLES used in this KEYFRAME image will be added to my next collection update ( sref.style ). #aivideos #sref #srefstyles #runway #gen3alpha #keyframetovideo #larssx

Try Runway Gen-3 Alpha to see how it compares to Sora. It is quite impressive at generating detailed, realistic video clips with high precision... #generativeAI https://lnkd.in/g8CE3STM

Gave a try to Runway Gen-3-Alpha Video-to-video. Converting some Quake speedrun to Apex Legend and some Mario to Rayman. Quite impressive overall! (Kept the jumping sounds in the quake footage :D ) #videogeneration #machinelearning #runwayml #videotovideo #video2video #deeplearning #generativeAI #generativevideo

Thoughts after using Runway for image to video ai gen (link at the end): 1. They did a good job of letting me know where I'm at with free credits. Paying customers are first-class users, everyone else is queued into oblivion. Kling was less up-front about this. They seem to have progress bars that aren't representing real progress. Kling still has me "minutes" away 2 days later. 2. My first attempt with Runway was my most successful. Kudos! 3. Prompt engineering is not intuitive. Despite asking for no camera movement, it still gave camera movement. On my second attempt, I asked for slow, lazy moving waves, and it gave me more waves than you'd expect in an hour long time lapse. It didn't just ignore what I asked for, it did the opposite. 4. The branded Runway watermark is not handled correctly. Baking this in for free videos is fine. BUT in paid accounts, TURN THIS OFF BY DEFAULT! My best video has your brand on it because I didn't know to turn this off. The watermark should be added in post, and you should always be able to serve up one with and one without the watermark. Don't bake it in unless you can do that. Cloudinary lets you do stuff like this easily. 5. Trying to rerun with the same prompt with watermark turned off was a waste of credits. The second time around the video generated was way off. It matters when you're basic plan gives users 5-6 chances per month. 6. Motion isn't quite there yet, even in the Gen-3 Alpha model, it is hit and miss. Water seems to be incredibly hard to get right. I had to blur the video to hide inaccuracies. It is still worth it at the current level of fidelity. Looking forward to what it can do in a year or two. 7. Using the same start and end frame doesn't guarantee a perfectly loopable video, but it should. For a 5-10 second video, I'd imagine many people want this feature. If you'd like to see what I was able to generate, visit https://luau.co #buildinpublic #startupjourney #generativeai #ai

The latest update to Gen-3 Alpha Turbo introduces advanced Camera Control features, enabling filmmakers to specify precise camera movements—such as horizontal, vertical, pan, tilt, zoom, and roll—directly within the AI video generation process. This enhancement offers unprecedented creative flexibility, allowing for dynamic and intentional storytelling through AI-generated content. While this update marks a significant advancement, further developments stil needed to make the film makers use it in their works. Integration with Industry-Standard Software - Seamless compatibility with tools like Adobe Premiere Pro or Final Cut Pro would streamline workflows, allowing filmmakers to incorporate AI-generated sequences effortlessly into their projects. Enhanced Motion Tracking - Incorporating advanced motion tracking could enable more complex camera movements and interactions between AI-generated elements and live-action footage, enhancing realism and coherence. Expanded Customization Options - Providing more granular control over camera parameters, such as focal length, depth of field, and motion blur, would allow filmmakers to achieve specific visual styles and effects. These potential enhancements could further position Runway's Gen-3 Alpha Turbo as an indispensable tool in the filmmaker's toolkit, bridging the gap between AI capabilities and traditional cinematic techniques. #AI #Filmmaking #Innovation #RunwayML #Gen3AlphaTurbo #visualcommunication #visualculture https://lnkd.in/gZdwFsHN

The age of AI is coming.

Mixing data with Unreal Engine Vered Camp, Beer sheva Building was captured using #laserscanning FARO Technologies #photogrammetry processed with Capturing Reality Environment scan using #gaussiansplats with Luma AI Give your scans context #nerf #realitycapture #virtualreality #3dscanning #heritage