Co-founder @MachgenAI. @Google, @Meta Alumn. Bay Area based. When not working on inference, you'll find me playing badminton or letting FSD drive.
Bay Area
Joined June 2013
- Tweets45
- Following179
- Followers45
- Likes41
Had a blast with this challenge. Pretty sure I won this round @KismatMachGen using @Kling_ai 4.0
Animated all the way
Round 2: @ManojMachGen, our CTO and co-founder, went 3D animated with @Kling_ai 4.0 โจ
Same script. Same music.
Match point: photorealistic or animated? ๐ธ๐
Manoj Krishnan retweeted
๐ฌ Made with MachGen Tuesday
The Galaxy Center by @simon_bare1837 : a ragtag band touring outer space, gig to gig, in their space van. A โ70s California road trip, in the cosmos. ๐โจ
#MadeWithMachGen
Manoj Krishnan retweeted
"Add the Transamerica Pyramid to the San Francisco skyline."
Edit one part only. One of our favorite features on @GeminiApp Nano Banana 2.1
๐ค Made with AI
Manoj Krishnan retweeted
@GeminiApp Nano Banana 2.1 is live on MachGen
โ Up to 14 reference images, 4 characters + 10 objects kept consistent
โ 1K, 2K or 4K, nine aspect ratios
โ Edit one part of an image, leave the rest untouched Faces and outfits finally stay the same edit to edit.
Manoj Krishnan retweeted
Love creating with MachGen AI? Become an affiliate and earn $10 for every friend who joins!
Share your link. When a friend signs up and tops up $10+, you both get $10 in credits.
Faster, cheaper AI video & images on top models.
Get your link ๐
Video made using @MiniMax_AI @ElevenLabs on MachGen
Manoj Krishnan retweeted
๐ถ๏ธ ๐ค โ A secret agent turned cowboy turned samurai
@Kling_ai 4.0 handles up to 10 keyframes and 15 references in one task: characters, videos, even voices.
Coming soon to MachGen, and we CAN'T wait to see what you'll create!
Stay tuned ๐
Kling 4.0 is coming this October. Kling 4.0 Flash is live now for Ultra Yearly subscribers.
Kling 4.0 brings video creation to a new level of visual realism, creative control, and narrative completeness.
๐๏ธ Upgraded Audio & Visuals: Stable dynamic motion, high-quality stereo audio, more accurate lip sync, up to 4K resolution, and 10-bit HDR output.
๐ฎ Omni Reference: Richer reference options with up to 15 multi-modal references, more consistent results, and enhanced video editing.
๐ฅ Seamless Storytelling: Multi-keyframe control supporting up to 10 keyframes, native 30-second generation.
๐ Diverse Possibilities: Video extension, support for multiple languages, accents, and dialects.
The stage is set. You call the shots!
Manoj Krishnan retweeted
Most inference infrastructure was built for LLMs. Diffusion came second.
So we built a stack for diffusion from scratch:
- Up to 6x faster video
- 2โ3x lower video cost Original, undistilled models.
Our CEO Kismat Singh shares more with @UniteAI and Antoine Tardif @futuristdotai ๐ unite.ai/kismat-singh-co-fouโฆ
Manoj Krishnan retweeted
MiniMax H3 was open-sourced two months ago and the @MachgenAI team has been busy building cool features on top of it.
H3 is no longer only one model but a collection of different pipelines & accelerated LoRAs. Sharing a few things that our engineering team built on top of Minimax-H3 ๐งต
Manoj Krishnan retweeted
Building audio apps usually means dealing with 30-second clip limits and messy third-party terms.
We run YuE2 on MachGen to bypass all of it. โก
โข Full commercial freedom (zero third-party terms)
โข 6-minute continuous tracks in a single API call
โข Exact verse-by-verse lyric execution
โข Scalable, developer-ready API
Stop fighting licensing traps. Build with YuE2 today ๐
machgen.ai/generate
Manoj Krishnan retweeted
๐ฆ๐ฒ๐ฒ๐ฑ๐ฎ๐ป๐ฐ๐ฒ 2.5 ๐ฃ๐ฟ๐ฒ๐๐ถ๐ฒ๐ is officially live on MachGen.
โก 480๐ฝ ๐๐ฟ๐ฎ๐ณ๐: Rapid, low-cost previews to test camera angles and pacing
โก 1080๐ฝ ๐๐ถ๐ป๐ฎ๐น: Render only your winning take in full high definition
Draft first. Refine later.
Try it on MachGen today.
The open-source community keeps pushing this space forward at an incredible pace. ๐งก
At MachGen AI, we are proud to offer Minimax H3 communityโs best innovations faster, cheaper, and accessible to everyone.
๐งก๐๐ฒ๐๐ ๐ฝ๐ฟ๐ถ๐ฐ๐ฒ๐ ๐๐ฉ๐๐ฅ ๐ผ๐ป ๐ ๐ถ๐ป๐ถ๐บ๐ฎ๐
๐๐ฏ - ๐ผ๐ป๐น๐ ๐ผ๐ป ๐ ๐ฎ๐ฐ๐ต๐๐ฒ๐ป, ๐๐ต๐ฟ๐ผ๐๐ด๐ต ๐ฆ๐ฒ๐ฝ๐ ๐ฏ๐ฌ!
R2V/I2V/T2V Minimax H3 768p: 0.04/sec
๐๐น๐๐ผ ๐น๐ฎ๐๐ป๐ฐ๐ต๐ถ๐ป๐ด ๐๐ฏ ๐ง๐๐ฟ๐ฏ๐ผ ๐ผ๐ป ๐ ๐ฎ๐ฐ๐ต๐๐ฒ๐ป. ๐ง๐ต๐ฒ๐ ๐ฎ๐ฟ๐ฒ ๐ฏ๐ฎ๐๐ฒ๐ฑ ๐ผ๐ป ๐๐ผ๐ฅ๐๐ ๐ฏ๐๐ถ๐น๐ ๐ฏ๐ ๐๐ต๐ฒ ๐ฐ๐ผ๐บ๐บ๐๐ป๐ถ๐๐
The first 500 users to top up $25 get H3 Turbo FREE through Sept 30!
R2V/I2V/T2V Minimax H3 Turbo 768p: 0.01/sec for others
๐จ๐๐ฒ ๐๐ฏ, ๐ณ๐ผ๐ฟ ๐ฏ๐ฒ๐๐ ๐พ๐๐ฎ๐น๐ถ๐๐.
Manoj Krishnan retweeted
๐งก๐๐ฒ๐๐ ๐ฝ๐ฟ๐ถ๐ฐ๐ฒ๐ ๐๐ฉ๐๐ฅ ๐ผ๐ป ๐ ๐ถ๐ป๐ถ๐บ๐ฎ๐
๐๐ฏ - ๐ผ๐ป๐น๐ ๐ผ๐ป ๐ ๐ฎ๐ฐ๐ต๐๐ฒ๐ป, ๐๐ต๐ฟ๐ผ๐๐ด๐ต ๐ฆ๐ฒ๐ฝ๐ ๐ฏ๐ฌ!
R2V/I2V/T2V Minimax H3 768p: 0.04/sec
๐๐น๐๐ผ ๐น๐ฎ๐๐ป๐ฐ๐ต๐ถ๐ป๐ด ๐๐ฏ ๐ง๐๐ฟ๐ฏ๐ผ ๐ผ๐ป ๐ ๐ฎ๐ฐ๐ต๐๐ฒ๐ป. ๐ง๐ต๐ฒ๐ ๐ฎ๐ฟ๐ฒ ๐ฏ๐ฎ๐๐ฒ๐ฑ ๐ผ๐ป ๐๐ผ๐ฅ๐๐ ๐ฏ๐๐ถ๐น๐ ๐ฏ๐ ๐๐ต๐ฒ ๐ฐ๐ผ๐บ๐บ๐๐ป๐ถ๐๐
The first 500 users to top up $25 get H3 Turbo FREE through Sept 30!
R2V/I2V/T2V Minimax H3 Turbo 768p: 0.01/sec for others
๐จ๐๐ฒ ๐๐ฏ, ๐ณ๐ผ๐ฟ ๐ฏ๐ฒ๐๐ ๐พ๐๐ฎ๐น๐ถ๐๐.
Manoj Krishnan retweeted
Advertising is entering its personalization era. The bottleneck was never creativity โ it was compute.
MachGen removes it: up to 6x faster, 2โ3x lower cost, with no quality tradeoff.
That takes an agency from one ad to 100 targeted variants by city, audience, or budget.
Trusted TV is already doing it in production: video generation down from minutes to ~10 seconds, with 10,000+ campaigns live.
Personalized, high-volume creative is becoming table stakes.
We're building the infrastructure underneath it.
Full conversation with Ian McCue โ youtu.be/nuKNgKl5hNg
Manoj Krishnan retweeted
Wan3.0 is now in beta on machgen.ai - built for shots that breathe, not clips that blink.
30 seconds. 1080p. 1 take.
Beta means two things here: you get it early, and you get it 20% off.
Try now: machgen.ai/models/alibaba/waโฆ
Manoj Krishnan retweeted
Seedance 2.5, live on day one. Whatever you pay, we pay.
Every closed model on MachGen runs at cost - zero markup, no subscriptions. Nobody undercuts zero. Our work is making models faster.
Try Seedance 2.5 now: machgen.ai/models/bytedance/โฆ
Manoj Krishnan retweeted
For years, the most beautiful videos a machine could dream of were closed and only lived behind an API.
You could rent the magic. You could never hold it.
This week, MiniMax H3 broke that wallโthe first open-weights video model to reach the edge of the frontier. But open weights that are slow and expensive are only half a promise.
So we kept the other half.
Today, H3 on MachGen generates a 5-second video an order of magnitude faster than your earlier experience, at a fraction of the cost.ย The weights are free. Now the speed is too.
Try here - machgen.ai/models/minimax/miโฆ
Manoj Krishnan retweeted
The biggest latency wins in AI inference don't come from the model itself. They come from the system around it. Here's why and how we built MachGen AI.
In our first founder blog, we share our bet on diffusion and world models, what's been built so far, and why we think this is one of the biggest opportunities in AI infrastructure.
โ machgen.ai/blog/founder/
Manoj Krishnan retweeted
Today weโre launching @MachGenAI out of stealth.
We build inference infrastructure purpose-built for image and video AI โ the diffusion and world models the visual internet is being built on.
The models raced ahead. The infrastructure underneath them didnโt. (๐งต)
Manoj Krishnan retweeted
Watching this animation is like a tour down memory lane of old processors.
Fascinating: Mooreโs Law predictions vs actual growth in transistor count.
by @datagrapha
reddit.com/r/dataisbeautifulโฆ
