Gradium Launches Free Voice Design to Challenge ElevenLabs
Paris-based startup Gradium has launched Voice Design, a free tool that generates synthetic voices from text prompts, outperforming rivals like ElevenLabs in blind quality tests.
Gradium, a voice AI startup spun out of the French research lab Kyutai, has released Voice Design. Available immediately in its Studio and API, the tool allows developers to generate entirely new synthetic voices using simple text descriptions specifying age, gender, pitch, and use case. Crucially, the feature is free across all tiers, including the free plan, positioning it as a direct, cost-free alternative to paid services from incumbents like ElevenLabs, Inworld, MiniMax, and Fish Audio. Because the voices are completely synthetic, developers can bypass the legal hurdles of voice cloning, such as actor licensing, consent agreements, and ongoing royalty payments.
In blind pairwise testing conducted by native speakers, Gradium's technology outperformed its competitors. Across 7,627 blind comparisons spanning five languages, Gradium won 72.6% of the matchups, placing 13.6 percentage points ahead of the next-best system. The tool showed particularly strong performance on regional accents, securing a 97% win rate for Québécois French, 86% for Rioplatense Spanish, and 85% for Bavarian German. An independent evaluation using Gemini 3.1 Pro as an LLM judge mirrored these results, scoring Gradium at 4.06 out of 5 compared to ElevenLabs at 3.86.
For developers, the API returns up to five candidate voices per prompt. Once a user selects and names a voice, the system provides a stable identifier that integrates directly into Gradium's streaming text-to-speech endpoint. However, because the underlying model relies on non-deterministic sampling rather than database retrieval, practitioners must save this identifier immediately, as the same prompt will not generate the exact same voice twice. The tool currently supports English, French, German, Spanish, and Portuguese, with over 100 regional accents.
The release comes nine months after Gradium emerged from stealth. Backed by $100 million in total seed funding from investors including Nvidia, the startup is led by co-founder Neil Zeghidour, a former researcher at Google Brain, DeepMind, and Facebook. By offering custom voice generation for free, Gradium aims to disrupt a market where competitors like ElevenLabs have achieved multi-billion-dollar valuations.
This is our own summary of reporting by AlphaSignal
