Gradium

Gradium is out of stealth to solve voice. We raised $70M and after only 3 months we’re releasing our transcription and synthesis products to power the next generation of voice AI.

@kwindla @covaldev 🚀

I care a lot about 1) latency, and 2) good benchmarks. The voice models from @GradiumAI occupy in the top spot on the @covaldev voice AI benchmark leaderboard, which tests both latency and accuracy. We've made the Gradium models the default in our open source, massively multi-player, LLM game. You play the game by talking to your "ship AI." Of course, humans aren't the only talkers in the universe, these days. So @chadbailey59 has been building OpenClaw bots that play the game, too! Here's a video of an OpenClaw agent playing the game by talking to the agent's ship AI. Chad created specific personalities for both the OpenClaw bot and the ship AI, here. Gradium's voice models are very emotive, and they support voice cloning and customization. I love how morose and pessimistic the ship AI is, and how excitable and cheerful the OpenClaw bot is. This is not scripted at all. These are two voice AI agents just doing their thing.

Gradium voices power conversations with the in-game ship AI as you trade, ally, and outmaneuver rivals. Real-time TTS in a game like this allows consistent character voices and clean turn-taking across very long, multi-session contexts.

You define them in code, run them in Vercel sandboxes, and plug them into the running game. The same pattern works for any live application that needs isolated, on-demand subagent execution alongside the main pipeline. producthunt.com/products/gradi…

You can now bring your own subagents to Gradient Bang. It's @pipecat_ai 's multiplayer, LLM-driven space trading game. Voice is the primary input, and Gradium is the default TTS model. Live on Product Hunt today

Link: luma.com/paris-hack

We're all set to meet the builders at the Paris AI Hackathon tomorrow.

Alright it’s now official - barely 9 months old and @GradiumAI is already trouncing the entire voice AI field on third party TTS benchmarks Better than OpenAI Better than Eleven Labs Better than Cartesia Better than Deepgram …

Most TTS benchmarks come from the vendors themselves. Conditions get picked to flatter the system being measured, using studio-quality text, simple inputs, and P50 only without spread The numbers are real. They're just not comparable. And they don't tell you what happens when 5% of your calls hit a 600ms tail. We built benchmarks.coval.ai to fix that. Standardized conditions, open methodology, open code, continuously updated. Today's snapshot: Gradium ranks first on both latency metrics we measure (P50 TTFA and P25-P75 spread) without trading off intelligibility. Their WER is competitive with the top of the field. Full per-provider breakdown is in our benchmarks page. Run it yourself: github.com/coval-ai/bench… Live results: benchmarks.coval.ai/tts

The 2ms P25-P75 spread is the number worth paying attention to. Nearly every request lands at the same latency. In production voice agents, that consistency is what makes turns feel natural rather than unpredictable. gradium.ai/blog/coval-tts…

Gradium's real-time TTS ranks best across all models on the latest benchmarks by @covaldev First in both latency and accuracy against ElevenLabs, Cartesia, Rime, Deepgram, and OpenAI. We wrote about the full results in our latest blog.

@tavily @nebius With @tavilyai and @nebiusai

Gradium is back in Berlin. We are hosting a hackathon with @Tavily and @Nebius on 29th May. This time focused on enterprise use-cases. Voice agents, search-augmented pipelines, or anything that needs real-time voice and external knowledge. Win prizes and credits.

Link to register: luma.com/juded1wb

I was in the flight when I got his idea. Here's another project I made during my break Audiobook ListenHER 💅 Turn any book you own (PDF), narrated in your favourite person's voice. Just get them to yap for few seconds and that's it! Made with @OpenAIDevs Codex Voice cloning from @GradiumAI

Voice cloned for demo purposes from 10 seconds of audio, sub-second turn-taking, and everything you see runs on Gradbot. github.com/gradium-ai/gra…

What is it like to have a consulting session with Darth Vader and pitch him a game? At Gradium, everyone creates their own voice agents to meet their specific needs. Here’s a Gradbot developed by our GTM expert, Timothee, who is passionate about gaming.

If you want to understand where the industry is at when it comes to TTS, STT and Speech to Speech, I would recommend watching the recent AI engineer video from @neilzegh He compares some of scenes from the "Her" movie which is so frequently floated around when it comes to the ideal "voice" experience and tells us what is still missing from that experience stack. Super glad that @aiDotEngineer exists @swyx :)

@optimizooor 🚀

Berlin did have game :) Back at a hackathon after a while and Big Berlin Hack felt like a new era: AI now multiplies what you can build in 1-2 days. I built Mira: an AI creative director that turns product briefs into launch videos, with voiceover powered by @GradiumAI / gradium.ai miracuts.com

@ClementDelangue @kyutai_labs @ElevenLabs Here you go. Made by gradbot in one command: "I want to create a vintage conversational agent using the talkie model huggingface.co/awilliamson/ta…"

who can help? maybe @GradiumAI @kyutai_labs @ElevenLabs?

Can someone build the Reachy Mini app that lets it speak like Talkie from @AlecRad in a posh pre-1931 British voice? I want to hear a next-gen robot channeling the 1920s! Space: huggingface.co/spaces/multimo… How to build a Reachy Mini app: huggingface.co/blog/clem/reac…

European Summer, a little bit different ;) If anyone ever needed an excuse to spend a weekend in summer in Paris, here it is! We are back, teaming up with OpenAI & Hexa with support from SLNG, Gradium, fal, XAnge & Fastino Labs for a 1-day 100-person hack in Paris in 2 weeks (16th of May). Applications are open now 👇 (our last Paris hack had >550 applications, so better be fast)