Video yükleniyor...

Video Yüklenemedi

Bu video yüklenirken bir sorun oluştu. Bu geçici bir ağ sorunundan kaynaklanıyor olabilir veya video kullanılamıyor olabilir.

Ana Sayfaya Dön

481 frames, 24fps 720 x 1280 resolution Total VRAM: ~40gb Rendered in ComfyUI using LTX 2.3 with audio + image as input. Still playing with settings but again... pretty good out of the box

A.I.Warper

36,936 subscribers

55,283 görüntüleme • 7 ay önce •via X (Twitter)

Oyun Sanat Bilim & Teknoloji

Anya Rossi• Live Now

Private livecam show

42 Yorum

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

Oh forgot Prompt executed in 117.28 seconds

L. Oscar Plains profil fotoğrafı

L. Oscar Plains7 ay önce

I can comfortably (for the most part) do 15 second clips at 1280x704 on my RTX 3060 with 32GB of system ram (+pagefile), it takes about ~12 minuets per run in total. Much slower than you, but still pretty neat I can do it with such a low-end card. Clip below is just text2video.

Linus ✦ Ekenstam profil fotoğrafı

Linus ✦ Ekenstam7 ay önce

H100 cloud gpu?

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

Yes

Salma profil fotoğrafı

Salma7 ay önce

really nice!! is this with audio conditioning?

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

Yeah just some random sample I have been testing with for years at this point. The voice is the streamer Valkyrae

Sharon | AI wonders profil fotoğrafı

Sharon | AI wonders7 ay önce

this is wan 2.2

RKA Scope profil fotoğrafı

RKA Scope7 ay önce

Awesome. What GPU did you use?

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

H100

JAMIE profil fotoğrafı

JAMIE7 ay önce

this is awesome

Brie Wensleydale🧀🐭 profil fotoğrafı

Brie Wensleydale🧀🐭7 ay önce

Thanks for running the test!~ Have you tried doing anything stylized or anime yet?

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

Not yet

Koburigo profil fotoğrafı

Koburigo7 ay önce

Tried something similar with the default LTX2.3 I2V workflow, and the output was absolute garbage every single time. Blurry wobbly shapes, nightmarishly inconsistent anatomy, background objects shifting around and disappearing, prompt was almost entirely ignored... How?

Sonia Tamayo profil fotoğrafı

Sonia Tamayo7 ay önce

really amazing lip sync & motions....

PUREVOODOO profil fotoğrafı

PUREVOODOO7 ay önce

Why did your video turn out so good quality? The videos I see are of poor quality and the mouth is distorted.

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

Great question no idea. I am straight rendering 720p (no upscaler) and running distil lora at 50% Euler + 8 steps

Jarnhand profil fotoğrafı

Jarnhand7 ay önce

@ComfyUI Can you PLEASE post prompt / wf. Mine looks like a horror show...

Tomasz profil fotoğrafı

Tomasz7 ay önce

1300 x 1300 24 fps, 24GB VRAM. 41 s LTX 2.3 Creation of modern Galatea.

Samuel Alvez profil fotoğrafı

Samuel Alvez7 ay önce

Would you mind sharing your workflow?

Saintus profil fotoğrafı

Saintus7 ay önce

@ComfyUI Was this image to video or video to video (body swap)

Tadhg Roche profil fotoğrafı

Tadhg Roche7 ay önce

Can we get a workflow 🥹

🔞🎨 Glitch Saint (0/2 comm slots) profil fotoğrafı

🔞🎨 Glitch Saint (0/2 comm slots)7 ay önce

Abysmal dogshit

John Koes profil fotoğrafı

John Koes7 ay önce

Really impressive result. I am using a 4090 and the results are not nearly as good.

DJ JD profil fotoğrafı

DJ JD6 ay önce

Probably a stupid question as im still feeling my way around comfy ui etc was this done on a cloud gpu like.runpod etc or done locally sorry if it seems a dumb question I only have 8gb vram so looking into running something cloud based

A.I.Warper profil fotoğrafı

A.I.Warper6 ay önce

Not a stupid question. I do not have much knowledge around optimizing workflows for lower end of the VRAM range these days so I’m unsure what exactly you could fit on that card with LTX 2.3

SebastianTorresVFX profil fotoğrafı

SebastianTorresVFX7 ay önce

Any idea if CN or Multimodal capabilities will come to comfyui for this model?

Berelz profil fotoğrafı

Berelz7 ay önce

DAaaaaaaaaaaaaaaaaaaam

Evan Emami profil fotoğrafı

Evan Emami7 ay önce

Porn industry got a free upgrade...

MattStopa profil fotoğrafı

MattStopa7 ay önce

Was this a straight shot? 20 seconds!? Curious how long it took to render.

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

120s No Sage attn, torch compile etc.. no speed ups

MattStopa profil fotoğrafı

MattStopa7 ay önce

Wow okay. That's nuts.

Mo's fav goat profil fotoğrafı

Mo's fav goat7 ay önce

Do you have a link for a tutorial how to do this?

Matija profil fotoğrafı

Matija7 ay önce

@ComfyUI does LTX support custom LoRAs?

Ctrl Alt You profil fotoğrafı

Ctrl Alt You7 ay önce

Yes indeed good job!

Diego P. Jaccottet profil fotoğrafı

Diego P. Jaccottet7 ay önce

@ComfyUI Quantized?

A.I.Warper profil fotoğrafı

A.I.Warper7 ay önce

@ComfyUI No full weights

God profil fotoğrafı

God4 ay önce

looks pretty good, what is you vram setup?40gb where can i find the workflow. thanks for sharing video

Diego IA profil fotoğrafı

Diego IA7 ay önce

Any oficial web? Is amazing results

Open Source AI profil fotoğrafı

Open Source AI7 ay önce

Great ! Workflow pleaz

whatsupfranks profil fotoğrafı

whatsupfranks7 ay önce

I wish you could use 2 RTX 5090s for 48gb of ram

py pa profil fotoğrafı

py pa7 ay önce

nice

Tharinda Marasingha profil fotoğrafı

Tharinda Marasingha6 ay önce

Generation time for this clip ?

Benzer Videolar

Anyways check out this 60s clip I made with a single image of Lex Fridman and a 20s audio recording of his voice. Full continuous shot with no cuts created with LongCat Avatar in ComfyUI. VibeVoice for the audio.

A.I.Warper

15,204 görüntüleme • 9 ay önce

zeroscope_v2 XL, A watermark-free Modelscope-based video model capable of generating high quality video at 1024 x 576 Model on Hugging Face : This model was trained with offset noise using 9,923 clips and 29,769 tagged frames at 24 frames, 1024x576 resolution. zeroscope_v2_XL is specifically designed for upscaling content made with zeroscope_v2_576w using vid2vid in the 1111 text2video extension by kabachuha. Leveraging this model as an upscaler allows for superior overall compositions at higher resolutions, permitting faster exploration in 576x320 (or 448x256) before transitioning to a high-resolution render. zeroscope_v2_XL uses 15.3gb of vram when rendering 30 frames at 1024x576

zeroscope_v2 XL, A watermark-free Modelscope-based video model capable of generating high quality video at 1024 x 576 Model on Hugging Face : This model was trained with offset noise using 9,923 clips and 29,769 tagged frames at 24 frames, 1024x576 resolution. zeroscope_v2_XL is specifically designed for upscaling content made with zeroscope_v2_576w using vid2vid in the 1111 text2video extension by kabachuha. Leveraging this model as an upscaler allows for superior overall compositions at higher resolutions, permitting faster exploration in 576x320 (or 448x256) before transitioning to a high-resolution render. zeroscope_v2_XL uses 15.3gb of vram when rendering 30 frames at 1024x576

AK

714,575 görüntüleme • 3 yıl önce

Wz Audio is much better since the update. I’ve been messing with the Asymmetrical audio frequency settings and have been using this with good results + Dolby Atmos. Obviously way more in-depth tunes out there but this has been working for me so far.

Wz Audio is much better since the update. I’ve been messing with the Asymmetrical audio frequency settings and have been using this with good results + Dolby Atmos. Obviously way more in-depth tunes out there but this has been working for me so far.

TeeP

261,141 görüntüleme • 1 yıl önce

Let's dance! Grok Imagine is so good with natural movements, keeping the style and the vibe of the input image. It blows my mind that it's still not even in a v1...

Let's dance! Grok Imagine is so good with natural movements, keeping the style and the vibe of the input image. It blows my mind that it's still not even in a v1...

Alex Patrascu

42,255 görüntüleme • 11 ay önce

Google Flow. Here’s the workflow: Image: Start with a "first-person perspective POV" prompt. Add specific visual style keywords like "early 2000s point and shoot" for a candid, textured look. Video: Set your image as the first frame using frames to video to lock perspective. Audio & Export: Prompt natural ambient audio, then export up to 4K. Check out the walkthrough to see it all in action.

Google Flow. Here’s the workflow: Image: Start with a "first-person perspective POV" prompt. Add specific visual style keywords like "early 2000s point and shoot" for a candid, textured look. Video: Set your image as the first frame using frames to video to lock perspective. Audio & Export: Prompt natural ambient audio, then export up to 4K. Check out the walkthrough to see it all in action.

Google Flow

31,996 görüntüleme • 27 gün önce

You might have seen some posts that Highguard runs at a fake in-game resolution, but its not exactly what you think. Steam User Captain_SmellyRat noticed this issue first by pointing out that if you set render scale at 100%,the game would still show 80% of the resolution in DLSS hud UI. So If you're playing at 1080p, the internal resolution would be 864p and If you don't change the default render scale (which seems to be always set to 75%) the final resolution would be 648p. However it turns out that the in-game resolution is changed in value directly from post-processing settings that you set. Post-processing quality settings would change the internal resolution as follows: Low - 80% of in-game resolution Medium - 90% of in-game resolution High - 100% of in-game resolution If you're playing at high settings, you will not notice the game looking blurry, but if you play at medium or low settings, the game will look extremely blurry even though you think you're playing at 1080p or 4k. I find it weird that post-processing quality directly affects the internal game resolution, but if you're playing the game and it looks blurry to you, make sure you set the post processing quality to high. Whenever this was intentional or not to achieve a higher FPS, I don't exactly know. However, I still think the developers should have clarified what each settings affects the game, which in this case here it directly affects the resolution.

You might have seen some posts that Highguard runs at a fake in-game resolution, but its not exactly what you think. Steam User Captain_SmellyRat noticed this issue first by pointing out that if you set render scale at 100%,the game would still show 80% of the resolution in DLSS hud UI. So If you're playing at 1080p, the internal resolution would be 864p and If you don't change the default render scale (which seems to be always set to 75%) the final resolution would be 648p. However it turns out that the in-game resolution is changed in value directly from post-processing settings that you set. Post-processing quality settings would change the internal resolution as follows: Low - 80% of in-game resolution Medium - 90% of in-game resolution High - 100% of in-game resolution If you're playing at high settings, you will not notice the game looking blurry, but if you play at medium or low settings, the game will look extremely blurry even though you think you're playing at 1080p or 4k. I find it weird that post-processing quality directly affects the internal game resolution, but if you're playing the game and it looks blurry to you, make sure you set the post processing quality to high. Whenever this was intentional or not to achieve a higher FPS, I don't exactly know. However, I still think the developers should have clarified what each settings affects the game, which in this case here it directly affects the resolution.

NikTek

20,832 görüntüleme • 8 ay önce

I'm sure you all are aware of the logo mapping glitch in #WWE2K26 but there's also a glitch with the image points themselves. Each image counts as 6pts currently regardless of resolution. CornellG It should be as follows: 512x512 = 6pts 256x256 = 5pts 128x128 = 4pts

Donquis29 👊🏾✊🏾💯

13,842 görüntüleme • 7 ay önce

Video clip of playing Beat Saber using the maxed out “Eureka mode” preset on Vision Pro (with PS VR2 Sense controllers) As a reminder, this should work with any OpenXR game out of the box TestFlight is awaiting approval. Join ClearXR discord for updates:

Video clip of playing Beat Saber using the maxed out “Eureka mode” preset on Vision Pro (with PS VR2 Sense controllers) As a reminder, this should work with any OpenXR game out of the box TestFlight is awaiting approval. Join ClearXR discord for updates:

Brad Lynch

30,792 görüntüleme • 6 ay önce

WanVideo has been lot of fun, currently playing with applying sliding context windowing similar to AnimateDiff, the 1.3B model is especially suitable for this due to it's speed, here's 1025 frames in one go, under 5GB VRAM used, but took about 10 mins on a 5090 with 30 steps.

WanVideo has been lot of fun, currently playing with applying sliding context windowing similar to AnimateDiff, the 1.3B model is especially suitable for this due to it's speed, here's 1025 frames in one go, under 5GB VRAM used, but took about 10 mins on a 5090 with 30 steps.

Jukka Seppänen

14,778 görüntüleme • 1 yıl önce

We used to buy more VRAM. Now we’re finding ways around the VRAM limit. This little DDR5 SO-DIMM setup is paired with an old NVIDIA Tesla P40 in the video. The P40 is an old datacenter GPU with 24GB of VRAM. By today’s standards, it’s not fast. But combine that with a machine carrying 128GB of system memory, and suddenly you have a very different local AI box. The GPU handles what it’s good at. The system memory gives the model somewhere to go when VRAM runs out. You obviously pay for that with bandwidth and inference speed. But this is the part I find interesting: Local AI is making “how much memory do I have?” almost as important as “how powerful is my GPU?” We’ve spent years chasing faster GPUs. Maybe the next wave of local AI hardware will be about finding smarter ways to use all the memory we already have. More VRAM is still better. But it’s no longer the only way to run a bigger model.

We used to buy more VRAM. Now we’re finding ways around the VRAM limit. This little DDR5 SO-DIMM setup is paired with an old NVIDIA Tesla P40 in the video. The P40 is an old datacenter GPU with 24GB of VRAM. By today’s standards, it’s not fast. But combine that with a machine carrying 128GB of system memory, and suddenly you have a very different local AI box. The GPU handles what it’s good at. The system memory gives the model somewhere to go when VRAM runs out. You obviously pay for that with bandwidth and inference speed. But this is the part I find interesting: Local AI is making “how much memory do I have?” almost as important as “how powerful is my GPU?” We’ve spent years chasing faster GPUs. Maybe the next wave of local AI hardware will be about finding smarter ways to use all the memory we already have. More VRAM is still better. But it’s no longer the only way to run a bigger model.

KyzoroX

35,554 görüntüleme • 28 gün önce

Run Stable Audio 3 on ANY Computer with NO VRAM 1-click launcher for the official Stable Audio 3 gradio app. 1. Cross Platform: Works on ALL OS (Mac, Linux, Windows) out of the box 2. No GPU Needed: Even runs on CPU. 3. Ultra Fast: Generate a 2 minute clip in split seconds

Run Stable Audio 3 on ANY Computer with NO VRAM 1-click launcher for the official Stable Audio 3 gradio app. 1. Cross Platform: Works on ALL OS (Mac, Linux, Windows) out of the box 2. No GPU Needed: Even runs on CPU. 3. Ultra Fast: Generate a 2 minute clip in split seconds

cocktail peanut

72,933 görüntüleme • 4 ay önce

🚨 INPUT-LAG TESTING RESULTS 🚨 🎯 EasyAim x Aimlabs x Kovaaks x AimBeast x AimingPro 🏆 (MONITOR GIVEAWAY INFO BELOW) Input-lag summary (in ms): - EasyAim: 2.11 - AimBeast: 3.14 - Kovaaks: 3.46 - AimingPro: 3.79 - Aimlabs: 5.48 In this video, we test input lag from mouse click to target disappearance. Results are at the end, along with detailed testing methods. Tests were run multiple times using lowest graphics settings, anti-aliasing off, and fastest settings to minimize input latency for each game. Try the test yourself independently if you have a monitor with Nvidia input-latency analyzer or LDAT. For any specific settings requests from Aimlabs KovaaK's Aimbeast Aiming.Pro, let us know, and we'll gladly re-test. 🎁 To enter the giveaway for a ROG Swift Pro PG248QP 540Hz monitor: - Like - Retweet - Follow 🍀 Winner announced on Friday, May 24th! Good luck everyone and let us know what do you think in the Tweets below! We listen to your feedback!

🚨 INPUT-LAG TESTING RESULTS 🚨 🎯 EasyAim x Aimlabs x Kovaaks x AimBeast x AimingPro 🏆 (MONITOR GIVEAWAY INFO BELOW) Input-lag summary (in ms): - EasyAim: 2.11 - AimBeast: 3.14 - Kovaaks: 3.46 - AimingPro: 3.79 - Aimlabs: 5.48 In this video, we test input lag from mouse click to target disappearance. Results are at the end, along with detailed testing methods. Tests were run multiple times using lowest graphics settings, anti-aliasing off, and fastest settings to minimize input latency for each game. Try the test yourself independently if you have a monitor with Nvidia input-latency analyzer or LDAT. For any specific settings requests from Aimlabs KovaaK's Aimbeast Aiming.Pro, let us know, and we'll gladly re-test. 🎁 To enter the giveaway for a ROG Swift Pro PG248QP 540Hz monitor: - Like - Retweet - Follow 🍀 Winner announced on Friday, May 24th! Good luck everyone and let us know what do you think in the Tweets below! We listen to your feedback!

EasyAim Trainer

57,020 görüntüleme • 2 yıl önce

Here's a look behind the scenes of how I used a 3d model, generated from an image, as a "driver" for ai animation. In the near future we'll see more 3d tools that support these types of workflows, with much more emphasis on enhancing user input instead of just letting the ai do everything. I used the 3d model to generate keyframes that I then animated in Luma Dream. By using a 3d model with a diffusion "layer" done with Krea, I got quite high quality frames with a relatively high consistency as well since the AI didn't have to "hallucinate" that much. I upscaled and detailed the frames using Magnific. Using semi-detailed 3d models as drivers for gen ai is very powerful, and it allows us to use models that are sculpted in a more free-flowing type of workflow that doesn't rely so much on high surface detailing or very time-consuming finish, but instead relies more on gestural 3d sculpting. #art #ai

Here's a look behind the scenes of how I used a 3d model, generated from an image, as a "driver" for ai animation. In the near future we'll see more 3d tools that support these types of workflows, with much more emphasis on enhancing user input instead of just letting the ai do everything. I used the 3d model to generate keyframes that I then animated in Luma Dream. By using a 3d model with a diffusion "layer" done with Krea, I got quite high quality frames with a relatively high consistency as well since the AI didn't have to "hallucinate" that much. I upscaled and detailed the frames using Magnific. Using semi-detailed 3d models as drivers for gen ai is very powerful, and it allows us to use models that are sculpted in a more free-flowing type of workflow that doesn't rely so much on high surface detailing or very time-consuming finish, but instead relies more on gestural 3d sculpting. #art #ai

Martin Nebelong

48,280 görüntüleme • 2 yıl önce

Vacuumed together with a bunch of balloons once again. It's been a while since I had the chance to do this, but gosh it's still as good as it always was

Vacuumed together with a bunch of balloons once again. It's been a while since I had the chance to do this, but gosh it's still as good as it always was

Blimped Sonder (for 983 days)

22,382 görüntüleme • 1 yıl önce

Borderlands 4 Performance Summary in 2 minutes: 0:00 - Native Res Badass Settings the game runs bad 0:30 - Upscaling Performance Mode Badass Settings, the game barely manages to hit 60FPS 1:20 - Frame Gen turned on and now you have 100 FPS Some people will turn all these settings on and be happy with it but in my honest opinion these are used as crutches to make the game look as if it performs good. Borderlands 4 has a certain art style that doesn't look like it needs these kind of settings to be turned on to play at a high framerate but sadly it does and you can defend this if you want but the future of gaming looks bleak, using this tech to not optimize the games further will just result in input latency increase as every game will be delivered with frame-gen turned on in the devs mind.

Borderlands 4 Performance Summary in 2 minutes: 0:00 - Native Res Badass Settings the game runs bad 0:30 - Upscaling Performance Mode Badass Settings, the game barely manages to hit 60FPS 1:20 - Frame Gen turned on and now you have 100 FPS Some people will turn all these settings on and be happy with it but in my honest opinion these are used as crutches to make the game look as if it performs good. Borderlands 4 has a certain art style that doesn't look like it needs these kind of settings to be turned on to play at a high framerate but sadly it does and you can defend this if you want but the future of gaming looks bleak, using this tech to not optimize the games further will just result in input latency increase as every game will be delivered with frame-gen turned on in the devs mind.

NikTek

359,686 görüntüleme • 1 yıl önce

literally the entire context of this whole sequence [its still so painful to watch with the way it frames Stolas as having a cutesy "falling in love all over again" moment whilst Blitzø's disgust and annoyance is played for laughs]

literally the entire context of this whole sequence [its still so painful to watch with the way it frames Stolas as having a cutesy "falling in love all over again" moment whilst Blitzø's disgust and annoyance is played for laughs]

Stolitz #1 hater & Chred Connoisseur

78,521 görüntüleme • 11 gün önce

Pushed another 2021 Unreal Engine experiment through LTX in ComfyUI! Not perfect, but look at those free waves! 🌊 Back in 2021, I was exploring the first version of Metahumans and ray tracing, trying to figure out how to generate water displacement. It’s crazy to see this extra layer of detail now on top of it. Although not all shots resulted in good waterflow direction . Combined a line-art + depth pre-processor to generate the shot guide. Also used these line-art generations in Nano Banana 2 to get the new first and last frames. Some shots are still tricky to get right if you want to keep control over the shot, especially the last shot with nothing in the end frame.

Pushed another 2021 Unreal Engine experiment through LTX in ComfyUI! Not perfect, but look at those free waves! 🌊 Back in 2021, I was exploring the first version of Metahumans and ray tracing, trying to figure out how to generate water displacement. It’s crazy to see this extra layer of detail now on top of it. Although not all shots resulted in good waterflow direction . Combined a line-art + depth pre-processor to generate the shot guide. Also used these line-art generations in Nano Banana 2 to get the new first and last frames. Some shots are still tricky to get right if you want to keep control over the shot, especially the last shot with nothing in the end frame.

Jerome | InsaneUnreal

20,275 görüntüleme • 7 ay önce

img2threejs 1.5.1 is out 🚀 Checkout: This release introduces a new approach to image → 3D → Three.js. Image → Tripo → img2threejs → reconstructed geometry → Three.js Instead of using the generated 3D model as the final asset, img2threejs uses it as a reference to rebuild the character into a representation that can be generated and rendered directly in Three.js. No Blender workflow, and no GLB required as the final output. The animation in the demo is still a bit rough because this release is focused on reconstruction, not animation. I’ll be improving that in the next release. Still experimental, but this is a big step toward where I want img2threejs to go.

img2threejs 1.5.1 is out 🚀 Checkout: This release introduces a new approach to image → 3D → Three.js. Image → Tripo → img2threejs → reconstructed geometry → Three.js Instead of using the generated 3D model as the final asset, img2threejs uses it as a reference to rebuild the character into a representation that can be generated and rendered directly in Three.js. No Blender workflow, and no GLB required as the final output. The animation in the demo is still a bit rough because this release is focused on reconstruction, not animation. I’ll be improving that in the next release. Still experimental, but this is a big step toward where I want img2threejs to go.

Nick

117,760 görüntüleme • 1 ay önce

Is it selfish to get air con in the heat wave? Air conditioner units are selling out as Brits struggle with the soaring temperatures, but with around 4% of total global greenhouse gas emissions attributed to air con, should we be using them when we know they're heating the planet?

Is it selfish to get air con in the heat wave? Air conditioner units are selling out as Brits struggle with the soaring temperatures, but with around 4% of total global greenhouse gas emissions attributed to air con, should we be using them when we know they're heating the planet?

Good Morning Britain

1,789,215 görüntüleme • 3 ay önce

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!

sharing an easy way to test DJev (Diffusion Gemma Jev), the next evolution of the incredible jev concept and fully open sourced (model, code, api...) imagine jev but it also has: - native image-input support - native image as options!! - real-time video streaming input - as fast as jev (not with images, 600ms) - same price as jev, but can get even cheaper - using Diffusion Gemma adapted - we are seeing on-par performance but happy to have external validation! It is a research preview made for fun, so don't expect stability and the awesome scaling as jev team has delivered! this was a weekend project :) have fun!

David Villalón

43,078 görüntüleme • 17 gün önce