Designer & creator. Building @EmotionEngineAI: stories, music, characters and creative tools. Mobile games since before they were cool.
Saved
Joined December 2018
- Tweets21.8K
- Following464
- Followers520
- Likes31.9K
Pinned Tweet
A little milestone for me.
I’ve been quietly building EmotionEngine for quite some time, and today I finally made the first public post.
Starting with music and with CLUB’d.
This is only the beginning. 🙂
All aboard. EmotionEngine.
I’m building EmotionEngine. A new creative system for giving AI-generated music, voices, stories and visuals stronger emotional structure, identity and the ability to move people.
Let’s kick things off with EE’s theme & test song:
“CLUB’d” — Lena — Live at Estadio Azteca, Mexico City
Mood: energetic, organic live concert
EE test: human–instrument hybrid timbres, instrument merges
Tools: Suno v5.5 guided by EmotionEngine.
No traditional prompt engineering, Suno Studio, DAW, filters, audio uploads or plug-ins used.
More EE experiments coming.
I made two small tests with the new Suno Speech (beta) vs. my EmotionEngine Speech v3.
I think EE delivered.
How much emotional shape can you give AI speech without scripting the performance?
Two EmotionEngine Speech v3 comparisons:
Suno Speech Beta — native
vs.
Suno 6 — guided by EE Speech v3
Hilla speaks about photography — moving through warmth, discovery, sadness and conviction.
Apostle John reads Genesis 1:1–5 in Greek, calm, restrained authority.
EE shapes the field around the performance: emotional geometry, timing, breath, vocal material, movement and parts of the background score, while leaving space for emergence.
No Suno Studio, DAW, plugins or post-production.
Same engine. Different emotional field.
🎧 Recommended
Maybe YouTube is the answer for serious content creators? They have had issues also, though...
This is just wrong.
Kalle Erkkilä retweeted
Aethr create is designed to get you the best out of musicgen and it's unlimited. We won't cap your downloads. As we had more models and Aethr Pro subscription will give you custom options including voice personas and deeper control of covers.
Also we are making our Mastering tool unlimited for everyone because music shouldn't feel limited.
Kalle Erkkilä retweeted
Today, we're sharing new research on Solaris, our first Interface World Model.
Solaris is a new kind of operating system that generates interactive interfaces frame by frame, in real time, with no code. We find that Solaris outperforms frontier LLMs when generating new interfaces, across structural similarity and information retention. Read more and request early access at the link below.
Kalle Erkkilä retweeted
Proud to say we now have over 1000 artists on AethrMusik
We are an artist-run platform for the artists. No labelling, no curveballs, just music and expression. We also thank the listeners of our platform and the apps are coming shortly <3
Kalle Erkkilä retweeted
You can start using Pika Soundtrack, Pika Music, Pika SFX, and Pika Speech exclusively on the Pika API Club:
dev.pika.art/models/pika/pik…
Kalle Erkkilä retweeted
MiniMax Music 3 is now live on our Open Source page
Free for everyone. We will also customise it and add our own features so let us know what you would like to see.
aethrmusik.com/opensource
🎵MiniMax-Music3
Next-Generation Open-Weights Production-Ready & Versatile Music Model
huggingface.co/MiniMaxAI/Min…
I asked ChatGPT to check out MiniMax Music 3. It knows my projects so it recommended it. But at the moment it's not available for mac, I guess.
I pasted MiniMax Music 3 links into Codex and told it to install the model, It set up the ComfyUI version, generated a test song, and built me a reusable personal plugin for it.
My gaming PC is the AI music service now. 👇
youtube.com/watch?v=-Ye71x8Y…
Kalle Erkkilä retweeted
Over 100 artist signs ups in one night
We just wanted to thank you all 💜
If theres any features you would like to see let us know. Our biggest upgrade to the site is happening over the course of a few weeks including Android and iOS apps
Kalle Erkkilä retweeted
Did Suno just KILL their AI Music Service????
This is really bad @suno
This new limit literally means that you can never download most of the songs you create!
Since you can not download them, you can't use them commercially either (apparently)!
You pay to create songs you can not use!!!
20 downloads per month is a total joke!!!
Will they charge users for more downloads?
Who owns the songs we don't download?
Please SHARE this as much as possible!!!!
I have to try this one. I took a look at their site. They are cooking.
I'm still working on a custom variant of the open source model based on Ace Step.
But for now I've created a free to use hosted version of Ace Step 1.5 with referencing and cover options. All it requires is an account on @AethrMusik which is free.
The idea is I put a little magic in it, but best results will be from using a reference track, it also masters with @AethrMusik and I think the results are decent personally but its something you can use for now. If under heavy load there is a queue system which won't be a thing with the custom version as I'll buy more server space.
Site: aethrmusik.com/opensource
Kalle Erkkilä retweeted
My GPU just went brr and made a 15 second video with native audio.
MiniMax H3 is imperfect, surprisingly capable, and already running on hardware far below a 5090. This is truly a wonderful AI summer for Open Source AI video peeps.
Check it: youtu.be/pPWnWcWNPrA
Kalle Erkkilä retweeted
Today, we’re announcing Bonsai 27B: the first 27B-class model to run on a phone.
Bonsai 27B is the new multimodal flagship of the Bonsai family. Based on Qwen3.6 27B, it brings a new capability tier to local AI: multi-step reasoning, structured tool use, long-context workflows, and coherent agentic loops.
Until now, models in this class have been impractical to deploy locally. A 27B model occupies roughly 54 GB in 16-bit precision, and even a strong 4-bit build is around 18GB - too large for a phone and for most laptops.
Bonsai 27B changes that.
It comes in two variants:
• Ternary Bonsai 27B: 5.9 GB, 1.71 effective bits per weight, optimized for laptop-class quality.
• 1-bit Bonsai 27B: 3.9 GB, 1.125 effective bits per weight, optimized for phone-class footprint.
Everything is open-sourced today under the Apache 2.0 license.
Kalle Erkkilä retweeted
Yes, Telegram won’t scan your private messages, no matter what banana-republic tricks the EU uses.
