Google DeepMind's Gemini Omni Flash has claimed the top Video Arena spot for both text-to-video and image-to-video generation, posting a 158-point Elo jump over Veo 3.1 at 1080p while also leading video-editing preferences. The native multimodal model accepts mixed text, image, audio, and video prompts to create or iteratively refine short clips, with early access already live for paid Gemini app users and YouTube Shorts creators.
🚀
Exciting news: Gemini Omni Flash is now #1 in the Video Arena (both Text-to-Video and Image-to-Video)!
For Text-to-Video this is a massive +158 pt improvement over Veo 3.1 (1080p) and a large +61 pt lead over the next best model, Seedance 2.0.
Congrats @GoogleDeepMind for this huge milestone!
Developers will get the model through the Gemini API and Agent Platform soon, but no concrete date, pricing, or rate limits have been shared yet.
Conversational multi-turn edits and multimodal references are called out as differentiators, though their edge outside the current human-preference benchmarks remains untested in public workflows.
Positive users hail Gemini Omni Flash topping video arena leaderboards as raising the bar for text-to-video and image-to-video, while negative users dismiss the results as fake benchmarks favoring Google over Seedance.
Gemini Omni Flash is SOTA at image to video, text to video, and video editing : )
Excited to get this to developers in the API soon!
Coming to API soon.
Exciting news: Gemini Omni Flash is now #1 in the Video Arena (both Text-to-Video and Image-to-Video)!
For Text-to-Video this is a massive +158 pt improvement over Veo 3.1 (1080p) and a large +61 pt lead over the next best model, Seedance 2.0.
Congrats @GoogleDeepMind for this huge milestone!
Check out the benchmarks page for Omni: https://deepmind.google/models/gemini-omni/#:~:text=how%20to%20prompt-,Performance,-Gemini%20Omni%20Flash
Gemini Omni Flash is SOTA at image to video, text to video, and video editing : )
Excited to get this to developers in the API soon!
📹 Omni is such a strong video generation and editing model, effectively @NanoBanana for video footage!
Coming to the API soon... 👀
Gemini Omni Flash is SOTA at image to video, text to video, and video editing : )
Excited to get this to developers in the API soon!
Exciting news: Gemini Omni Flash is now #1 in the Video Arena (both Text-to-Video and Image-to-Video)!
For Text-to-Video this is a massive +158 pt improvement over Veo 3.1 (1080p) and a large +61 pt lead over the next best model, Seedance 2.0.
Congrats @GoogleDeepMind for this huge milestone!
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 🧵
> OpenAI trying to shift the responsibility to Deepmind my dawg is washed…
@OfficialLoganK @GoogleDeepMind Brah got be real with you folks you guys need to step up and deliver a mythos class model. Side quests like this is irrelevant
Gemini Omni Flash won 82% of its head-to-head battles in Battle Mode (without ties). Incredibly high compared to the next best model’s win-rate of 69%.
@OfficialLoganK @miketaylorcai @GoogleDeepMind Can’t wait to interact with it
Gemini Omni Flash is SOTA at image to video, text to video, and video editing : )
Excited to get this to developers in the API soon!
Gemini Omni Flash also ties for #1 in Image-to-Video. Here we see a +77 pt improvement over Veo 3.1.
@OfficialLoganK @GoogleDeepMind Google Korea B2C "ME"
@OfficialLoganK @GoogleDeepMind Seedance users : sure bro.
@OfficialLoganK @GoogleDeepMind Brah got be real with you folks you guys need to step up and deliver a mythos class model. Side quests like this is irrelevant
Check out official Gemini Omni Flash release here: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-omni/
@OfficialLoganK @GoogleDeepMind Anyone who is using Seedance 2 knows these scores are ridiculous
Omni Flash is first and foremost about new capabilities like editing and multimodal references, but it's also a significant improvement over our previous models on raw text-to-video.
I hope you find the model useful. Let us know what we should improve.
Exciting news: Gemini Omni Flash is now #1 in the Video Arena (both Text-to-Video and Image-to-Video)!
For Text-to-Video this is a massive +158 pt improvement over Veo 3.1 (1080p) and a large +61 pt lead over the next best model, Seedance 2.0.
Congrats @GoogleDeepMind for this huge milestone!
The rate of progress in video generation remains mind blowing 🚀🚀🚀
Exciting news: Gemini Omni Flash is now #1 in the Video Arena (both Text-to-Video and Image-to-Video)!
For Text-to-Video this is a massive +158 pt improvement over Veo 3.1 (1080p) and a large +61 pt lead over the next best model, Seedance 2.0.
Congrats @GoogleDeepMind for this huge milestone!
@OfficialLoganK @GoogleDeepMind My next Strategy!!💎🎌 ⬇️Analyze more
@OfficialLoganK @GoogleDeepMind Yes please! Got any idea on price? Per second or something
Dive into the Video Arena leaderboard details at: http://arena.ai/leaderboard/text-to-video
@OfficialLoganK @shlomifruchter @GoogleDeepMind Hey @grok what is HappyHorse and who made it?
