Apparently SeeDance 2 is down for the time being... on BytePlus but here's a young wizard with his dragon companion. It's a very cool video model. This video cost $0.72 USD to generate which seems reasonable. (300,000 tokens for a 15 secs at $0.0024 USD per 1000 tokens.)show more

Rumor has it Dreamina is adding a new member... to the Seedance family. Dreamina Seedance 2.0 mini reportedly launching June 15 could deliver quality close to Seedance 2.0 at nearly half the cost. If true, this changes the game for creators who love Seedance but feel the burn on video generation costs. High-quality AI video is getting cheaper. That's great news for everyone building with it. A model that costs less than half the price doesn't just change the spec sheet it changes how often you actually hit "generate." 🔗 Looking forward to what Dreamina announces next. #dreamina #dreaminaseedance2mini #AIVideoshow more

📢📢📢Introducing xGen-MM-Vid (BLIP-3-Video)! This highly efficient multimodal language model... is laser-focused on video understanding. Compared to other models, xGen-MM-Vid represents a video with a fraction of the visual tokens (e.g., 32 vs. 4608 tokens). Paper: Website: Researcher’s 🧵:👇show more

$CAT is now listed on Binance! Here's a quick... FAQ for the community: Q: Where can I trade Simon's Cat on Binance? A: You can trade CAT on Binance here: Q: Why is it listed as #1000CAT? A: This format was adopted to align the token's price more closely with conventional trading values on Binance. For example, if $CAT's price on DEXs is $0.00007, it will trade at $0.07 on Binance, denoted as #1000CAT. Q: Will my tokens reduce when I deposit them to Binance? A: No, your tokens will be scaled down by 1000 for trading purposes on Binance, but will return to their original amount when withdrawn. For example, depositing 1000 $CAT will show as 1 #1000CAT on Binance, and withdrawing 1 #1000CAT will convert back to 1000 $CAT tokens in your wallet. Q: Does this mean Binance is issuing a new token? A: No, it's the same $CAT token as before. This change only simplifies trading on the platform but everything else remains the same. As we move towards achieving a new milestone in this amazing Simon's Cat journey, we welcome everyone to trade on Binance WITH MUCH MORE STILL TO COME, STAY TUNED!show more

AI video is about to get a lot cheaper.... I'm hearing Dreamina Seedance 2.0 mini, expected to launch on June 15, could deliver near Seedance 2.0 quality at dramatically lower cost. If true, this isn't just another model launch. It's the democratization of top-tier AI video generation. The biggest advantage in AI video has never been talent. It's access. And that gap may be about to disappear. #dreamina #seedance #dreaminaseedance2minishow more

JD Vance on Disneyland being shut down for his... family vacation: “We had the island to ourselves which was very cool. I had never been to Disneyland. I thought it was awesome. …sorry to all the people that are Disneyland for the longer lines, but we had a very good time.”show more

we just cut AI video prices in HALF on... Revid Ultra (Seedance 2.5): 200 → 100 credits per 5s Seedance 4K: 1000 → 500 Pro: 60 → 50 why: models got cheaper for us, price follows. wild concept apparently 👀 everyone is selling "unlimited" and making it impossible to generate with queue, throthling, rate limits ... we bill per second and auto-refund failed generations at 50% off now -> double the Seedance 2.5 footage live on Revid 🙌show more

The main hurdle holding back AI video generation right... now is no longer the fidelity. It is the expense. Word is going around about a fresh launch dubbed Dreamina Seedance 2.0 mini, said to be dropping on June 15, which promises to deliver premium Seedance-tier footage for a fraction of the cost. Assuming the output rivals Seedance 2.0, we are looking at top-tier video production becoming available to a dramatically wider pool of artists. Cheaper generation translates directly to bolder experiments, wilder creativity, and more founders shipping products. This is absolutely a development you need to monitor. #dreamina #dreaminaseedance2minishow more

#EscapeFromTarkov ** PEACEKEEPER BUG / PSA ** Seems like... this isn't affecting everybody but it's definitely happening to me so I thought I'd share. Buying ANY ammo it seems from Peacekeeper zaps all your USD - I would avoid buying ANYTHING from him at this time to be safe. You can move your spare USD to your PMC if you need to buy some stuff, however, if you have a lot, you will not be able to do this due to the USD limit on your PMC.show more

Reports are circulating of $AVAX tokens being airdropped into... wallets that participated in Memecoin Rush. ⚠️ Remember to please exercise caution and stay safe. Scammers will take any opportunity they see, in scenarios such as this ⚠️ Trader Joe is not responsible for the airdrop....But here is a cool video of Volly airdropping a giant red triangle.show more

The true bottleneck in the AI video scene right... now is no longer visual quality. The real issue is the cost. Word on the timeline is that the upcoming Dreamina Seedance 2.0 mini release is going to deliver top-tier video generation at a drastically lower price point. If the output fidelity is anywhere near the core Seedance 2.0 model, this will instantly democratize premium AI video for a massive wave of global creators. Slashing the cost barrier directly translates to relentless experimentation, explosive creativity, and a whole new ecosystem of builders. This is absolutely a major launch to keep on your radar. #dreamina #dreaminaseedance2minishow more

NVIDIA just released a very impressive text-to-video paper. Video... Latent Diffusion Models (Video LDMs) use a diffusion model in a compressed latent space to generate high-resolution videos. Here's a brief overview of how it works: 1. Pre-train image LDM on a dataset of images. 2. Turn the image LDM into a Video LDM by adding temporal layers to model video frames. 3. Fine-tune the Video LDM on encoded video sequences to create a video generator. 4. Temporally align diffusion model upsamplers to generate high-resolution videos. 5. Validate Video LDM on real driving videos of 512x1024 resolution, achieving state-of-the-art performance. 6. Apply the approach in creative content creation with text-to-video modeling. Paper: Project:show more

Sora 2 is great, but can it play Doom?... Well yes, apparently it can. Here's a 🧵with 10 games generated by @openAI's Sora 2. Drop your suggestions for games to try and I'll reply back with a video.show more

"So it's finally come to this, has it?" Witness... the ultimate anime transformation! From a quiet student in class to a winged warrior with a fiery sword, brought to life with Seedance 2 in MartiniArt. Follow, Retweet and Comment "seedance". I will DM you 500 credits for Seedance 2.0 and my exact workflow for this video. #MartiniArt #Seedance2 #AIAnimation #TextToVideoshow more

Most recent diffusion language model research (that I’ve seen)... seems to be using masking as the noising process. It looks like, however, most closed-source models (Google Gemini Diffusion and possibly Inception Labs’ Mercury) use a different noising process, where instead of masking tokens, they replace them with different tokens (either with a random token or a semantically similar token). I wondered how they were getting such high throughput with the latter noising process, since I believed that optimizing inference with KVCache approximation would be more difficult (for various reasons). I visualized this noising process with tiny-diffusion and compared it to normal unmasking, and was very surprised to see how fast the generation “settles” into a reasonable output, and then only slightly refines afterwards, requiring much fewer steps in total. Unmasking (where tokens are never remasked, the typical implementation) is inherently limited in generation speed by the fact that an increase in tokens decoded per step leads to more errors due to the mismatch between individual and marginal token probability distributions we sample from. The token replacement noising process seems to have a much different set of characteristics. Because we sample each token per step, every token makes “progress” towards the final output each iteration (in addition to *potentially* giving other tokens more information in future steps). Generally, masking has outperformed other noising processes, which is probably why most research focused on it (using smaller models). But the paper referred to in the retweet shows that random replacement as a noising process may scale better as model size increases. Big labs might have noticed these results much earlier (due to having drastically more training resources and being able to test larger models), which may explain the discrepancy in the choice of noising process. I’m gonna test this with larger models, since tiny-diffusion only has 10M parameters.show more

I saw this video on YouTube a month back... but IDGAS about it. But Now a video is going viral on Twitter in which a family bought a goat for sacrifice on the occasion of #BakraEid. But his neighbors are protesting and chanting hanuman chalisa to stop him from performing the Qurbanishow more

Fuck yeah! MaskGCT - New open SoTA Text to... Speech model! 🔥 > Zero-shot voice cloning > Emotional TTS > Trained on 100K hours of data > Long form synthesis > Variable speed synthesis > Bilingual - Chinese & English > Available on Hugging Face Fully non-autoregressive architecture: > Stage 1: Predicts semantic tokens from text, using tokens extracted from a speech self-supervised learning (SSL) model > Stage 2: Predicts acoustic tokens conditioned on the semantic tokens. Synthesised: "Would you guys personally like to have a fake fireplace, an electric one, in your house? Or would you rather have a real fireplace? Let me know down below. Okay everybody, that's all for today's video and I hope you guys learned a bunch of furniture vocabulary!" TTS scene keeps getting lit! 🐐show more