Browse · People
Youssef Rizk
Wondercraft cofounder whose work on AI audio emphasizes complete creator workflows, controllable speech and access to existing content.
Youssef Rizk cofounded Wondercraft with Dimitris Nikolaou, building an AI podcast creation platform around script generation, synthetic voices and publishing tools. In his September 2023 Latent Space interview, he described its mission as “expanding access to content”: turning existing material into formats and languages more people could consume. His argument for working at the application layer centered on serving a particular user’s whole workflow and choosing underlying speech engines to fit it.[2][1]
Before Wondercraft, Rizk worked at Palantir and pursued an earlier startup, Moonshot. That history provides context for his account of choosing a technical product and testing whether customers would pay before developing it further.[2][1]
Earlier career and the Moonshot pivot
Rizk grew up in Egypt, where he said he spent his first 18 years. He obtained an MEng in electrical and electronic engineering from Imperial College London, then spent four years as a software engineer at Palantir. There he led its startup practice and met Nikolaou. Rizk described forward deployed engineering as combining coding, solutions architecture, contract review, sales and customer interaction; he regarded that breadth as useful preparation for starting a company.[1][2]
Moonshot offered investment in athletes’ future success. Recalling the venture in September 2023, Rizk characterized its central challenge as legal and regulatory rather than technical. He said the founders concluded that they were building something people did not want and working outside their strengths. They decided to pursue something technical, ranked possible ideas and selected the content-conversion idea that became Wondercraft.[2][1]
Rizk and his collaborators tested the idea with a minimal React form without authentication. After someone shared it on Twitter and inbound interest grew, they improvised a Stripe payment flow that sent paying users to a separate application. Rizk recalled running on one EC2 instance. The immediate question was whether anyone would pay before they proceeded: he advised founders considering a pivot to try to invalidate an idea quickly rather than become attached to it.[1]
Building the creator workflow
By September 2023, Wondercraft used GPT-3/4 for text and ElevenLabs for voices, with tools around pronunciation, editing, music, default introductions and endings, hosting and publication. Rizk argued that this surrounding workflow mattered because general-purpose APIs served many possible uses without providing a complete product for each one. His advice was direct: “Focus on the use case, and then build a product for it.” This was his case for application-layer defensibility, rather than a demonstrated economic moat.[1]
In the same interview, Rizk announced a Zapier integration that could pick up newsletter content from an RSS feed and turn it into a podcast episode. He explained that automation could stop at generation: users could review and modify the episode before publishing, or schedule publication for later.[1]
For dubbing, Rizk distinguished speaker-turn alignment from lip synchronization. Wondercraft aligned translated speech to the chunks in which each person spoke, so a dubbed voice would match the person on camera. The implementation he described did not synchronize lip movements. He also said professional translators participated in dubbing quality assurance, addressing the difficulty of judging natural speech and accents in languages a customer did not understand.[1]
Voice quality, disclosure and spending
Rizk treated voice quality as dependent on the material. The team chose Anna for Hacker News Recap because she sounded professional, formal and neutral; conversational audio, he argued, benefited from pauses, stutters, laughter and overlapping speakers. He evaluated speech systems for realism and reliable control, including the ability to specify a tone or meet a promised segment duration. A 30-second segment was his example of a production requirement, not a reported service guarantee.[1]
Rizk defended natural-sounding synthetic speech as a way to make listening pleasant and advised creators to disclose AI-generated content. He warned that undisclosed synthesis could be interpreted as deception and provoke backlash. In the September 2023 discussion of watermarking, he identified a shared, interpretable standard as the obstacle to adoption.[1]
His spending principle was “GPUs after PMF”—product-market fit. Rizk warned that buying GPU capacity could substantially shorten a startup’s runway, while allowing that the investment could make sense at the right time or for a company developing its own text-to-speech technology.[1]
The 2025 EnterpriseAM Egypt Forum identified Rizk as Wondercraft’s cofounder and CTO.[3]
From the Latent Space archive
Original posts and separately reviewed explanations.
Sources
- 1Heralds of the AI Content Flippening — with Youssef Rizk of Wondercraft.aiAccessed 2026-09-18
- 2Wondercraft: Create video by chatting to AI | Y CombinatorAccessed 2026-09-18
- 3EnterpriseAM Egypt ForumAccessed 2026-09-18
Model source and teaching reviews passed. Human editorial review pending. This profile describes the sources as of 2026-09-17.