Today, we’re excited to introduce Miso One, the most emotive voice model in the world.
Miso One is an 8-billion-parameter text-to-speech model for highly expressive speech generation. It emotes like a human and responds faster than a human, with just 110 milliseconds of latency.
We’ve open-sourced the model weights, with API access coming soon.
Hear how Miso One sounds in the thread below.
This is where Livepeer finds it’s fit. Pulte Homes and SpaceX are putting GPUs everywhere, making their own networks. AI gateways are now common wheels anyone can build decentralized network with. Centralized providers are facing operational cost limits https://t.co/9ipSxsX795
🚨 Sam Altman warns OpenAi and Anthropic are experiencing severe pullback on Ai spending as companies put significant restraints on spending to restrict costs. The company warns investors it’s the first time this has happened in Ai and something we never expected. The buildout costs aren’t sustainable to allow profitability to hyperscalers or end users.
$soxx $dram
This is impressive work from @RyanOnTheInside and the @DaydreamLiveAI team @Livepeer. We've open sourced the DEMON repo which can run locally on your 3090/4090/5090 GPU, or you can try the hosted web version at https://t.co/ISKf8nwNSU.
@DaydreamLiveAI This in combination with a tool I’m building for Ableton to control it and build tracks with voice would be like peas and carrots. Can’t wait to get into it.
True real-time parameter control unlocks a wave of opportunities:
- Live remixing and performance: Instantly restyle and shape tracks.
- Responsive audio: Dynamically generate audio in installations, games, apps
- Style exploration: Audition new genres with music style LoRas
Introducing DEMON: a real-time, local-first diffusion engine for continuous music generation and control
Running at 25Hz with sub-100ms e2e latency on a single consumer GPU, it turns generative audio into an interactive streaming surface built to be played like an instrument.
NEW RELEASE Scope v0.2.5 from @DaydreamLiveAI
Current image-to-agent avatars land as flat talking faces. Scope v0.2.5 sharpens conversational avatar workflows by skipping CPU offload on high-VRAM GPUs, boosting generation speed.
Did you know you can save money by connecting Claude CLI to @BlueClawNetwork for your agent coding tasks? Get the step-by-step guide here https://t.co/Kzgdkt4EqZ
"Draw me a highly detailed Livepeer Real-Time Pipelines Star Wars-inspired space opera canteen, but instead of smugglers it’s packed with protocol engineers, orchestrators, video devs, and AI creatives all hacking on live AI video. Make sure to make it funny with in‑crowd jokes."
Quick #Scope tip from one of our engineers: in the Workflow Builder, you can wire multiple sinks into the same pipeline and compare outputs side by side, live.
Feed the source to the pipeline, split the output two ways: One branch through rife, one "raw".
Take the same source and pipeline and get two easily comparable outputs.
Works the same with upscalers or community-built nodes.
More Scope tips incoming!
Hey, new Scope 0.2.3 is out, and here's what's worth pausing on: A/V model improvements, running on the @Livepeer network, billing and credits system in place, and more...)
The clip below shows LTX-2.3 generating video and audio in a single pass, staying locked end-to-end throughout the session. That wasn't a given before this release.
Sharing more details in our thread 🧵
I vibe coded a particle canvas, and put it through @DaydreamLiveAI Scope to give it texture in real time. It's just a simple example of using AI to make realtime visuals.
If you are interested, come to my talk at the Austin TD meetup tomorrow!
https://t.co/J8Ip7Yenpw
.@iab: 2026 is the "defining moment" for AI-powered video operating in real time.
Real time is the key word here. Not batch processing or render queues.
Infra optimized for sustained, low-latency video workloads wins this.
Soon your claude/openclaw agent can use the Livepeer network to generate images all night long if you want, and soon - be able to pay for it over bankless, trustless escrow
Claude Code + Nano Banana 2 is f*cking cracked 🤯
I built a system inside Claude Code that researches any brand, writes 40 ad prompts from scratch, and fires them all to Nano Banana 2.
One brand name + one URL = 40 production-ready static ads.
All inside Claude Code.
I took @alexgoughcooper's brilliant framework and automated the whole thing inside Claude Code.
Perfect for DTC brands and agencies who need high-volume ad creative without briefing a designer or spending hours in Canva.
If you're finding winning ad concepts on Meta and manually recreating them one at a time in Higgsfield — copying prompts, pasting product details, tweaking aspect ratios, downloading, organizing...
This system eliminates the entire loop:
→ Give Claude a brand name and URL
→ It researches the brand's fonts, colors, packaging, and photography style
→ Builds a Brand DNA document from scratch
→ Fills in Alex's 40 proven ad templates (headline, us vs them, testimonial, UGC, review cards, stat callouts) with brand-specific details
→ Fires every prompt to Nano Banana 2 with your product photos as reference
→ Downloads finished ads into organized folders with an HTML gallery
No Higgsfield.
No manual prompt filling.
No copy-pasting between tools.
What you get:
→ 40 ad formats filled with your exact brand colors, fonts, and copy
→ 4 variations per format so you pick the best output
→ Product photos passed as reference so the model matches your real packaging
→ A reusable system — new brand, new folder, same pipeline
Built 100% in Claude Code with Nano Banana 2.
I put together a full playbook & Loom video showing the exact process to set this up yourself.
Want access for free?
> Like this post
> Comment "NANO"
And I'll send it over (must be following so I can DM)