Paris-based AI voice startup Gradium launched from stealth on December 2, 2025, with a $70 million seed round led by FirstMark Capital and Eurazeo. The company emerged from French nonprofit AI lab Kyutai and is building developer infrastructure for real-time speech-to-text, text-to-speech, voice agents and voice cloning.
The funding is unusually large for a company founded only months earlier. It signals strong investor confidence in voice interfaces—but it does not, by itself, prove that Gradium has superior technology, meaningful revenue or product-market fit.
As an Amazon Associate I earn from qualifying purchases.
What happened with Gradium?
Gradium announced its launch and $70 million seed financing on December 2, 2025. FirstMark Capital and Eurazeo led the round. Disclosed participants included Xavier Niel, DST Global Partners and Eric Schmidt, among other investors.
Gradium was founded in September 2025 by Neil Zeghidour, a Kyutai founding member and former Google DeepMind voice-model researcher. No valuation was disclosed, and the public announcement does not establish whether the round included any earlier financing beyond the announced seed.
#1 Best Overall
- Stay present in every scenario: Every conversation is covered, in person, on calls, and online. 4 MEMS + 1 VPU microphones with AI beamforming capture every voice across the room. Smart Dual-Mode Recording switches automatically between phone calls and in-person. The free Plaud Desktop captures online meetings without a bot
- Walk out of every meeting with notes ready to act on: Plaud Intelligence transcribes in 112 languages with speaker labels and turns each recording into action items, decisions, and follow-ups, structured and ready to use. Choose from 10,000+ customizable templates tailored to your role and industry
- AI summary ready before you reach your desk: Auto Transfer moves each recording to the Plaud app automatically, and AutoFlow transcribes and summarizes so your notes are ready before you are back at your desk. Upgrade anytime to Pro (1,200 min/mo) or Unlimited
- Access your AI workspace anywhere: One connected workspace across Plaud Desktop, Plaud Web, and the Plaud mobile app, so your conversations and finished work follow you everywhere
- Your conversations stay private and yours: Compliant with ISO 27001, ISO 27701, SOC 2, HIPAA, GDPR, and EN 18031, with zero data used to train AI models. Trusted by 2.5M+ professionals, including legal, medical, and business professionals handling sensitive information
The size of the round is notable because Gradium was only a few months old at launch. It should be read as evidence of investor conviction about the market and founding team—not as proof of commercial traction. Publicly reviewed materials do not provide revenue, retention, named-customer, uptime or independently verified performance figures.
TechCrunch reported the funding, launch date, investors and Kyutai connection.
What Gradium sells
Gradium is positioning itself as voice infrastructure for developers rather than primarily as a consumer voice assistant. Its documented product includes:
- Streaming text-to-speech (TTS).
- Streaming speech-to-text (STT).
- REST APIs for non-streaming requests.
- WebSocket APIs for real-time audio.
- A library of prebuilt voices.
- Instant and higher-fidelity Pro voice cloning.
- Pronunciation dictionaries.
- Controls such as speed, temperature and voice similarity.
- Translation features and audio formats suited to real-time and telephony applications.
The current API documentation lists https://api.gradium.ai/api as the REST base URL and wss://api.gradium.ai/api for WebSocket connections. API access is authenticated with an API key. Developers can use the platform as one component in a larger voice-agent system; they may still need separate tools for reasoning, orchestration, telephony, monitoring and safety.
Gradium’s likely use cases include customer-care systems, voice agents, gaming characters, language-learning products, advertising and other conversational interfaces. The company’s broader vision is to make voice a primary interface for AI systems, but the public documentation is currently centered on STT and TTS capabilities.
Gradium’s API reference documents its authentication, REST and WebSocket interfaces.
The Kyutai connection matters—but needs careful interpretation
Gradium’s strongest differentiating story is its lineage. The company was spun out of Kyutai, a French nonprofit AI research lab associated with Moshi, an open-source full-duplex voice model.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →That connection gives Gradium access to a credible audio-research pedigree. Its founders came from Kyutai’s audio-research group, and the company says its proximity to the lab can help it commercialize frontier generative-audio research.
However, Kyutai’s open research and Gradium’s proprietary commercial production systems should not be treated as the same thing. Public materials reviewed for this article do not fully detail the organizations’ governance, intellectual-property ownership, licensing, research access or exclusivity arrangements. They also do not establish whether Gradium’s production models are directly derived from Moshi or simply share personnel and research lineage.
That distinction matters to customers and investors. A research association can help recruit talent and accelerate development, but long-term differentiation will depend on Gradium’s own models, infrastructure, reliability, distribution and customer economics.
Who founded Gradium?
Neil Zeghidour is identified as Gradium’s founder and chief executive. He previously worked on voice models at Google DeepMind and was a founding member of Kyutai.
Free tools Windows power users keep installed
One-click scans. No signup required.
Rank #2
- AI-POWERED TRANSCRIPTION & SUMMARIES: Plaud Note Pro is your professional voice transcriber, delivering high-accuracy transcription in 112 languages with auto speaker labels. Powered by top AI models and thousands of templates, Note Pro instantly creates structured summaries, mind maps, To-Do lists, and proposals tailored to your role and industry
- ENHANCED CONTEXT WITH MULTIMODAL INPUT: Capture audio, type notes, add images, and press to highlight key moments for richer context. During recording, instantly mark key moments with a single button press. Simultaneously enrich your audio by snapping photos of important documents or typing in ideas
- CHAT WITH YOUR RECORDINGS USING "ASK Plaud": Unlock deeper insights with this interactive AI. Ask questions, extract key points, draft emails, and get next-step suggestions—all grounded in your original audio for reliable, ready-to-use answers
- INTELLIGENT RECORDING WITH AI DIRECTIONAL AUDIO: Enjoy seamless, intelligent recording with Plaud Note Pro. Its AI automatically switches between call and meeting modes while recording, while directional audio and real-time spatial awareness minimize noise to capture voices with crystal clarity
- Everything Included: Includes Plaud Note Pro, magnetic case, magnetic ring, charging cable, and a free Starter Plan with 300 transcription minutes per month. Upgrade anytime in the Plaud app to Pro Plan (1,200 min/mo) or Unlimited Plan(Up to 24 hours of transcription per user per day)
Secondary company research identifies a broader founding team consisting of:
- Neil Zeghidour, CEO
- Olivier Teboul, CTO
- Laurent Mazaré, chief coding officer
- Alexandre Défossez, chief science officer
The broader team list comes from secondary research rather than the main launch coverage, so it should be treated accordingly. The more important point is that Gradium combines startup leadership with experience in speech and audio-model research.
How fast is Gradium?
Gradium says its expected streaming time-to-first-token is below 300 milliseconds. This is a potentially important target for voice applications, where delays can make conversations feel unnatural.
But time-to-first-token is not the same as end-to-end conversational latency. A user’s perceived delay can also include:
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →- Network round-trip time.
- Voice-activity detection and turn-end detection.
- Time to produce the first audible audio sample.
- Buffering and playback.
- Agent reasoning and tool calls.
- Interruption handling and cancellation.
As a result, Gradium’s below-300-millisecond figure should be attributed to the company and should not be presented as an independently verified conversational benchmark or as proof that Gradium is the fastest provider.
For a voice agent, the practical questions are whether a user can interrupt naturally, whether the system stops speaking promptly, how quickly it resumes, and how it performs with background noise, echo or multiple speakers. The reviewed sources do not provide independent comparative testing of those conditions.
Gradium’s documentation lists its supported languages and stated streaming latency target.
Languages and voices
Gradium’s current documentation lists five supported languages:
- English
- French
- German
- Spanish
- Portuguese
The public voice library contains 237 voices: 100 in English, 43 in French, 50 in German, 20 in Spanish and 24 in Portuguese. These are current product figures documented after the December 2025 launch, not necessarily the figures available on launch day.
Five languages give Gradium a potentially useful European focus, particularly for products serving several major European markets. They do not amount to broad global language coverage, and voice count alone does not establish equivalent quality across languages, accents, emotional range or difficult acoustic conditions.
The documented voice catalog provides the current language breakdown.
Rank #3
- YOUR AI PERSONAL ASSISTANT FOR EVERYDAY PRODUCTIVITY: More than a voice recorder, Pocket works as your AI personal assistant to capture, transcribe, and summarize meetings, calls, and ideas instantly. Core features are included out of the box, with optional advanced tools available for power users.
- ONE-TAP RECORDING FOR REAL-LIFE MOMENTS: Capture meetings, phone calls, and in-person conversations instantly with a simple tap, no typing, no interruptions, just effortless note-taking anywhere you go.
- SMART AI INSIGHTS & ORGANIZATION: Pocket automatically turns recordings into clear summaries, key action items and structured conversation maps so you can quickly review what matters without digging through audio.
- TURN CONVERSATIONS INTO ACTION WITH “ASK POCKET”: Don’t just record, understand. Instantly ask questions across your meetings, extract key insights and generate next steps in seconds. All grounded in your recordings, so answers stay accurate and reliable.
- MAGSAFE COMPATIBLE FOR SEAMLESS USE: Easily attach Pocket to your iPhone or other MagSafe compatible devices for convenient, hands-free recording on the go. Perfect for capturing meetings, calls, and ideas without needing to hold your device.
How voice cloning works
Gradium documents two voice-cloning options:
- Instant Voice Clone: built from approximately 10 seconds of reference audio.
- Pro Voice Clone: requires at least 30 minutes of clean audio, with about two hours recommended for stronger emotional range and stability.
Gradium says explicit consent from the voice owner is required. That requirement is important, but technical access to cloning does not resolve the wider legal and ethical questions. A commercial deployment may also require performer agreements, publicity-rights clearance, disclosure policies, content provenance and controls against impersonation or fraud.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
A short audio sample may make experimentation easy, but buyers should not assume that a technically convincing clone is automatically cleared for commercial use.
Gradium documents its custom-voice creation requirements.
Who is Gradium competing with?
Gradium enters a crowded market that includes ElevenLabs, Cartesia, Deepgram, OpenAI, major cloud providers, open-source speech models and specialist tools available through platforms such as Hugging Face. Anthropic, Meta and Mistral are also relevant to the broader AI-agent and speech ecosystem, even when they are not direct substitutes for every Gradium feature.
The useful comparison depends on the buyer’s problem:
| Buyer need | What to compare |
|---|---|
| Real-time voice agents | Streaming latency, turn detection, interruption handling, WebSocket quality and integrations |
| Narration and media | Voice quality, emotional control, language coverage and licensing |
| Contact centers | Transcription accuracy, telephony codecs, concurrency, compliance, reliability and support |
| Voice cloning | Consent controls, clone fidelity, moderation and rights management |
| European deployments | Data location, contracts, language coverage, support and regulatory posture |
| Enterprise infrastructure | Service-level agreements, security reviews, private deployment and volume pricing |
There is no current apples-to-apples benchmark in the reviewed material showing that Gradium beats these alternatives on latency, accuracy, price or uptime. A buyer should test the exact workload rather than choose based on a funding headline or a general “low latency” claim.
What the $70 million proves—and what it does not
The financing suggests that prominent investors believe voice interfaces could become an important layer of AI applications. It also reflects confidence in the founding team and in the possibility that Kyutai’s research reputation can help Gradium recruit talent and win early customers.
AI infrastructure companies can require substantial capital for model training, inference hardware, engineering, sales and reliability. A large seed can therefore provide meaningful time to build before revenue becomes substantial.
It does not establish:
- That Gradium has achieved product-market fit.
- That its voice quality or latency is superior to competitors.
- That it has significant recurring revenue.
- That it has a durable customer base.
- That its models are cheaper to operate at scale.
- That the company has disclosed a valuation attractive to outside investors.
The launch announcement said Gradium was onboarding early adopters in gaming, AI agents, customer care and language learning. That indicates commercial interest, but the reviewed sources do not provide enough information to quantify revenue, usage, retention or customer concentration.
Recommended Free Tools
Gradium’s launch press release describes its positioning and early-adopter focus.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Current product economics and limits
As documented on August 18, 2026, Gradium offered a free plan, paid plans priced at $13 per month, $43 per month, $340 per month and $1,615 per month, plus custom enterprise pricing. The free plan is marked as non-commercial, so a successful prototype may need a paid plan before it can be deployed for business use.
Rank #4
- Plaud Intelligence: Capture conversations in 112 languages and generate accurate transcripts with the Plaud App and Web. Plaud Intelligence uses leading models like GPT-5.5, Claude Sonnet 4.6, and Gemini 3.1 Pro to transform raw audio into structured insights. Choose from over 10,000 professional templates to generate mind maps and to-do lists, turning hours of discussion into immediate clarity
- Multiple Ways To Wear With Included Accessories: Adapt Plaud NotePin S to any workflow instantly with four included accessories. Wear your device effortlessly as a necklace, wristband, clip, or pin. Plaud NotePin S features a dedicated physical record button for precise, tactile control. Stay professional and keep your intelligence within reach all day
- Enterprise-grade Privacy: Built to the highest standards with ISO 27001/27701, SOC 2, HIPAA, GDPR, and EN18031 compliance. Every conversation is secure and protected. It is the trusted choice for creative, medical, and business professionals handling sensitive info
- Multimodal Input & Multidimensional Summaries: Capture audio, type notes, add images, and press/tap to highlight for richer context with multimodal input. Press the record button to mark key moments in real time. Plaud transforms a single conversation into multiple perspectives, providing faster, clearer insights, and unifies these inputs to deliver role-specific summaries that reflect your intent and priorities
- Lightweight Power and Peace of Mind: Weighing only 0.61 oz, Plaud NotePin S delivers 20 hours of continuous recording and 40 days of standby time. Store up to 64GB of audio locally, ensuring you capture every insight even without an internet connection
Usage is metered in credits:
- TTS: one credit per character.
- STT: three credits per second.
- STT translation: four credits per second.
- Speech-to-speech translation: 30 credits per second.
The FAQ lists sessions of up to 300 seconds and a 1,500-character per-session limit on the free tier. These details make workload modeling important. A product using long-form synthesis, continuous transcription, translation or speech-to-speech conversion can consume credits at very different rates.
Teams should model characters, audio seconds, translation type, concurrency, failed requests, retries and session length—not simply estimate cost by hours of audio.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Scan for outdated or missing drivers - takes under a minute3Repair Windows errors before they cause bigger problemsGradium’s pricing page lists its plans and credit rates. Product prices and limits can change, so customers should confirm the current terms before committing.
Practical buyer assessment
Gradium may appeal to developers that want:
- A specialist API for streaming speech.
- English, French, German, Spanish and Portuguese support.
- A European company and research lineage.
- Voice cloning and pronunciation controls.
- WebSocket access for real-time applications.
- A platform that can be combined with an existing agent or telephony stack.
It may be a weaker fit for teams that require:
- A long public production and uptime history.
- Independently validated latency comparisons.
- Broad worldwide language coverage.
- Native phone-number, SIP or complete contact-center operations.
- On-premises deployment.
- Established enterprise service-level agreements and support processes.
Before adopting the service, a serious evaluation should test noisy audio, accents, interruptions, barge-in, simultaneous speakers, long sessions, pronunciation edge cases, concurrency, failure recovery and credit consumption. Voice-cloning projects should additionally document consent and rights clearance before any customer-facing launch.
Key risks and unanswered questions
Latency claims may not translate into better conversations
A low streaming time-to-first-token can coexist with slow turn-taking if detection, network, buffering or agent reasoning adds delay.
The Kyutai relationship is not fully public
The research connection is valuable, but public materials do not fully explain corporate structure, intellectual-property rights, licensing or exclusivity.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallFive languages are useful but limited
Gradium’s language set may fit European use cases while remaining too narrow for globally distributed products.
Cloning creates rights and safety obligations
Consent is a stated requirement, but customers remain responsible for verifying rights, preventing misuse and determining when synthetic speech should be disclosed.
Funding is not traction
The size of the round and the reputation of its investors do not replace evidence about revenue, retention, production reliability or customer outcomes.
Bottom line
Gradium is notable because a prominent French audio-research lineage has become a heavily funded commercial voice-infrastructure company. Its documented APIs, multilingual launch, streaming architecture and cloning tools make it a credible platform to evaluate for voice-agent and speech applications.
The investment case is less settled. Gradium’s below-300-millisecond figure is a company target rather than an independent benchmark, and public materials do not yet establish superior quality, scale, revenue or reliability. The central question is whether its Kyutai-derived research credibility can become dependable, differentiated infrastructure that customers use at high volume.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
