The Streets of AI

@icreatelife Who Survived? 🧟 After They Woke (AI Short Zombie Film) youtu.be/Wh0grjGZUck

Post an AI video you took time to create and feeling proud about

@mark_k But rest of the world this week as well? Or do we have to wait further weeks? Or further months?

@TheStreetsOfAI Probably China first

Seedance 2.5 release is imminent from ByteDance

@IndieGameJoe This is just a modern version of Marco’s Magic Football on the mega drive … still good though

This indie dev is making a game where you kick a soccer ball on your way to school through beautiful yet challenging levels - Unlock new levels - Expand your room - Don't be late for school It's called Kick. Would you play this?

@youraigirl24 Mini is really bad. Every gen I made had many errors. I then spent more credits having to redo my video using Fast … which is much better

🚨 The wait is over: Dreamina Seedance 2.0 mini is officially live on Dreamina today. And to be completely honest? We might be looking at the most crucial AI video release of the entire year. It is not because it tries to be the heaviest, most powerful model out there. Instead, it completely flips the economics of generating AI video. Here is exactly why you need to pay attention 👇

@Safwon_CoderAI Mini is terrible. Every gen I tried was awful. You then spend more credits re trying with Fast (which is great by the way)

🚨 AI video finally reached its "GPT-3.5 moment." Dreamina Seedance 2.0 mini has officially dropped today. But I feel like the majority of folks are misinterpreting this drop. This goes beyond just launching another generation model. It's a clear indicator that the AI video space is starting a whole new era. Let's dive into why 👇

@seeksteve It’s not so much the voices .. it’s random noises during the video. So it’s not like you just need to switch up with voice acting .. you need to deal with random noises at any point during the gen

@TheStreetsOfAI sounds problematic. but maybe not problematic if you’re editing out the audio (a lot like the current version of grok imagine)

Everybody saying the new Google Omni model sucks compared to Seedance are right. It sucks for the things it’s not intended to replicate. btw: the masses don’t care about your fight scene or AI superhero storyboard, everyone scrolls past those. Omni excels at exactly what most AI creators are sleeping on right now. Real world use cases: - Video editing - Local understanding (like my example below) - Explainer visuals - Educational content This model will be responsible for bringing the masses into AI video because the next group adopting this technology doesn’t care about making the fight scene. They care about: - Not getting fired when their boss tells them to go learn that AI stuff - Making videos that the customers of their client won’t cringe at because it looks AI - Staying on brand for their business content in a way that doesn’t take hours. Here’s a local use case made with a slight remix to @fofrAI’s prompt: “Make a rapid fire video that shows a different neighborhood of Charleston SC every 1s, upbeat music, include text to label it.” All things considered on a very basic prompt, this video nails it.

@mpusgaje @JunieLauX That’s devastating 😞 How can we have regressed? We were teased with the future and it was snatched away! Omni Flash is a broken mess 😢

@TheStreetsOfAI @JunieLauX Nope. Sora is untouchable. The best approach now must be using ai voice at elevenlabs and put it in the video prompt

@IqraSaifiii It’s embarrassing. Most gens don’t get through the ridiculous guardrails. Occasionally one gets through and has clicking noises in the audio…. And only 10 seconds! I miss Sora 2, which was so much better

Gemini Omni Flash is so Bad As an ai film maker, I feel handicapped using this model , it feels like I am back in 2024. Inconsistent, no character consistency No memory, bad prompt following Write what you guys felt this is my opinion

@Lentils80 It’s dreadful. Most gens are against guardrails for no reason … the few that get through have terrible audio. Sora 2 was better than this mess in every way

@Boluwatifeolad7 @GoogleDeepMind It’s broken though. The output has clicking noises.. it can’t be used! Plus, nearly every prompt has to be tried 5x to get past crazy guardrails

@TheStreetsOfAI @GoogleDeepMind 😄 I'm not disappointed because unlike most, I'm not expecting a new video model. You all thought this is new but it's still running on the same old veo3 model. The only thing here is the new 10 seconds duration and the omni feature where you could use various elements as input.

We’re dropping Gemini Omni: our first step towards a model that can create anything from anything - starting with video. It combines Gemini’s intelligence with our generative media systems - representing a leap forward in world understanding, multimodality, and editing 🧵

@chrisnk14 @GoogleDeepMind Tell me you haven’t used Omni Flash, without telling me you haven’t used Omni Flash. It’s a complete mess!

@GoogleDeepMind Google will win the AI race. I am still convinced.

@Boluwatifeolad7 @GoogleDeepMind Have you actually used it? 🤣 It’s dreadful! You can’t make anything that could pass as even average.

@GoogleDeepMind Finally, something from GOOGLE🍹 Welcome back to the race. You left the chat a while ago, glad to have you back. LFG🔥

@GoogleDeepMind You should be embarrassed! The sound is terrible! Popping in every video. Guardrails are insane.. Only 10 seconds! What on earth have you been doing for a year? I miss Sora 2 so bad

Google Omni Flash is AWFUL! The sound is terrible, every video has popping sounds. The guard rails are ridiculous! You can’t extend clips. Only 10 seconds! Google should be embarrassed. I miss Sora 2 so bad!! #OmniFlash #Veo4 #GoogleOmni

@Waguri_Kaoruko8 Do you think there will be a pro version? 10 seconds isn’t enough

New VEO vs Seedance 2 I don't think this is the final version, but still my expectations were too far from being met. What do you say?

@cheatyyyy @synthwavedd @Lentils80 It doesn’t look any better than Veo 3.1 It doesn’t seem any better at consistency. It doesn’t seem any better sound. And only 10 second length? Much worse than Sora and that’s had time to live and die before Omni releases 🤣

@synthwavedd @Lentils80 it's a decently close alternative to seedance that will be available for pretty cheap on the gemini app so why not? less action focused but def not dead on arrival we're probably missing demos of actual peak capability, it's a lot better than veo for sure, just wish it was 15s

A first preview of Google's upcoming new video model, Omni, courtesy of @Lentils80! Unfortunately it seems dead on arrival versus Seedance 2. lol

@mark_k @khaledawad_98 @GoogleDeepMind But Omni seems to be capped at 10 seconds and not impressive at all! Is there a chance there could be Veo 4 within Flow? Omni seems incredibly disappointing for a year’s worth of work.

@khaledawad_98 @GoogleDeepMind Yes you can!

Gemini Omni is coming, a new omnimodal video model from @GoogleDeepMind. As opposed to Veo, the Omni model would be deeply integrated with Gemini itself, just like Nano Banana is. This could give it much better prompt understanding and reasoning skills. 🔥