Super intelligence for film and ad visuals. Super intelligence video workflows. DMs are open. Founder of Massive Studios
Joined March 2008
- Tweets34.3K
- Following945
- Followers39.9K
- Likes41.9K
Pinned Tweet
Krea 3 Ultra just landed at #4 on the Artificial Analysis image editing leaderboard.
1151 Elo. Nano Banana 2.1 is at 1152.
That's just one point.
Weird to see an indie lab in that lineup, huh? 😁
Style refs and LoRA training are next!
🤝 Paid partnership
today, Krea 3 makes it to the top of image arenas.
- "Fast" pushes the pareto curve of speed and quality.
- "Ultra" ranks top-4, only surpassed by OpenAI models.
the only company valued under $1T in the top-10!
more below 👇
My synth adventures continue...
This time, with a new synth from scratch: X-WAVE 80
Still WIP, but damn, it's fun making these!
And yes, it's still running 100% in the browser.
Gemini 4 Argon is a beast of a model:
The best way to explore product design ideas with Pinterest 👇
Add 3–4 references to Krea 3, describe your idea in a few words, and watch the magic happen.
Give each reference a purpose: product shape, materials, color palette, brand identity, lighting, or environment.
Combine a packaging reference, a material study, and an architectural photograph to imagine an entirely new product.
That’s what I explored here: fictional brands with their own names, visual identities, believable materials, and carefully considered environments.
Different ingredients open up different directions—from sculptural objects to products you could almost imagine buying.
What impressed me was how much direction you could communicate with a very short prompt.
The ingredients carry a lot of the detail, leaving you free to focus on the idea and quickly explore variations.
For this Product showcase, I generated 30 concepts and selected ten favorites.
Then Codex helped take them through the full production:
- Subtle six-second animations using Seedance 2.5 via the Krea MCP.
- An editable Tesseract composition with floating, bordered images and ingredient stacks that open like a fan.
- Carefully timed entrances, reversed exits, source credits, and the actual generation prompts.
- A one-minute edit over a soft deep house soundtrack made with ElevenMusic v2.5
Very fun, fast, and intelligent model. It went straight into my default toolbox.
Good job @viccpoes and @asciidiego 👏
Holy shit...
I think Higgsfield discovered a gold mine here with Flux 3 upscaling.
This has so many applications! Imagine how much archival footage we can restore.
Just watch this 👇
🤝 Paid partnership
FLUX 3 Upscaler from @bfl_ai is live on Higgsfield.
> 50% faster and 50% cheaper than before
> Upscale and sharpen any video
> FLUX 3 Video now generates in native 2K and 4K
Available now, with 7 days of exclusive access on Higgsfield.
Just when you think image models can't get any better...
Krea 3 shows up.
The quality is wild. Look at this:
I saw this question on X:
"Based on your extensive knowledge and training data, what is the world's single greatest unconventional life hack?"
So I asked Claude, ChatGPT, Gemini, and Grok Bot:
This is awesome, but damn, they make it REALLY hard to try it out...
It boggles my mind how they make things so hard to use.
OmniHuman 2.0 is live.
It doesn't just sync lips — it understands your audio. Feed it any track, speech, music or dialogue, and it reads the tone, emotion, and every word, then performs it. Every syllable lands.
Now generate up to 5 minutes of video in a single continuous shot, up from 60 seconds, with no stitching required.
And 𝟰𝟬%+ 𝗰𝗵𝗲𝗮𝗽𝗲𝗿 — $0.12 → $0.07/sec. Better quality, lower cost.
Deeper understanding. Longer takes. Lower price.
byteplus.com/en/product/Omni…
20K+ unique visitors in the last 28 hours.
On a project that wasn't even supposed to be something in the first place.
X is the best platform to launch a product.
Nothing compares to it, and you can clearly see that.
I can't believe this worked...
Gemini 4 Argon made a Jupiter/Moog combo synth in under 30 minutes.
It's all in a single HTML file, and you can play with it here: argonsynth.netlify.app/
Somebody asked me in the comments if I could do a simplified kids' version of the Argon Synth.
So, I went into Antigravity and got to work.
Here's Argon Junior:
argonsynth.si/junior/
I can't believe this worked...
Gemini 4 Argon made a Jupiter/Moog combo synth in under 30 minutes.
It's all in a single HTML file, and you can play with it here: argonsynth.netlify.app/
I can't believe I did this in less than a day...
Gemini 4 is incredible.
Argon Poly-8 is now a fully functional VST plugin 👇
I can't believe this worked...
Gemini 4 Argon made a Jupiter/Moog combo synth in under 30 minutes.
It's all in a single HTML file, and you can play with it here: argonsynth.netlify.app/
I always show my friends AI videos of me, and I always get the same reaction:
"Dude, this is great... but the sound is just off. It doesn't sound real. If someone solves that, it's game over."
Well, HeyGen just did 👇
🤝 Paid partnership
The numbers from their blind tests back it up:
- People rated their own digital twin 76/100, ahead of Cartesia (63) and ElevenLabs v4 (59)
- Users picked it as their own voice over ElevenLabs v4 62.2% of the time
- It beat Cartesia 87% of the time when the voice switched languages
- It's #1 on @ArtificialAnlys
Try it now: heygen.com/
It's live on the main HeyGen platform starting today.
F*CK Yeah!
This is immense. It will save me so much time.
Opus 5.5 and GPT-6 Astra can now color grade your videos at the level of a professional.
Introducing color grading with Tesseract.
Shift individual colors, enhance your camera log footage, and dial in the look shot by shot with support for ACES and LUTs.
Every layer stays editable, whether you work in your agent or take your project into Premiere or After Effects.
Tools, apps, and models used:
Generation: Higgsfield @higgsfield_ai
Seedance 2.5, Nano Banana Pro.
I used the Higgsfield CLI to submit jobs and the Higgsfield MCP connector for uploads and the project.
Claude: @claudeai
Claude Code desktop app, with Opus 5.5 as director, editor, and producer.
Audio analysis (local, RTX 5090):
Demucs htdemucs_ft split the song into stems; htdemucs isolated the vocals in each rap clip.
beat_this found the beats and downbeats.
faster-whisper large-v3 transcribed the lyrics.
torchaudio MMS_FA timed each word.
librosa measured key, onsets and feel.
The AST AudioSet model classified the audio.
Picture prep and sync (local):
BiRefNet made the mattes (type behind Jolly, plus keep-out boxes for his face and chain).
MediaPipe tracked faces and poses.
The skill’s own sync_plate.py measured word-level lip-sync lag.
Edit, type and render:
The music-video skill’s Node render engine (napi-rs/canvas, fontkit, bidi-js).
Anton font (Google Fonts, OFL licence).
ffmpeg and ffprobe for decoding, H.264 encoding, BT.709 tagging and the AAC mux.
A Python 3.11 venv for the scripts (numpy, soundfile and the rest).
Updated again 👇
The synth now has a mobile-friendly version.
Big update to ARGON!
Browser-based 80s analog poly synth workstation.
• Chord Progression Tracker
• Dedicated Sub-Bass Tracker + Sidechain
• 10-Track Analog Beat Engine
• Track Mutes
• Record & Export to WAV & MP3
Available at argonsynth.si
Big update to ARGON!
Browser-based 80s analog poly synth workstation.
• Chord Progression Tracker
• Dedicated Sub-Bass Tracker + Sidechain
• 10-Track Analog Beat Engine
• Track Mutes
• Record & Export to WAV & MP3
Available at argonsynth.si
I can't believe this worked...
Gemini 4 Argon made a Jupiter/Moog combo synth in under 30 minutes.
It's all in a single HTML file, and you can play with it here: argonsynth.netlify.app/
Alex Patrascu retweeted
I can't believe this worked...
Gemini 4 Argon made a Jupiter/Moog combo synth in under 30 minutes.
It's all in a single HTML file, and you can play with it here: argonsynth.netlify.app/