tiny-voca: 18 MB text-to-speech model runs in the browser
tiny-voca is a lightweight English TTS model trained only on synthetic speech and built for $9.56.
A compact text-to-speech model called tiny-voca has appeared. It was trained only on synthetic speech, but the result is quite good.
The male and female voices were created and voiced by Qwen3-TTS. The model weighs 18 MB and runs in a browser on any device. For now, it only speaks English.
The entire development cost the author $9.56 — less than a pizza.
The demo can be tested for free here, while the model itself is available on Hugging Face.
Why it matters
AnalysisIt shows that extremely small TTS models can deliver usable voice output without heavy infrastructure, making audio generation more accessible.
Discuss in community
Share your questions and insights with developers
Related in Ecosystem
Practical guides, core terminology, and agent skills
Modern Speech Synthesis (Text-to-Speech / TTS)
The technology of artificial generation of human speech from printed text. Modern neural network TTS models (ElevenLabs, OpenAI Audio, Chatterbox) reproduce natural intonations, logical stresses, breathing, timbre, and emotional coloring, indistinguishable from a live narrator.
Jev and System One AI: A Complete Practical Guide to the Decision Model from TypeSafe
A deep engineering breakdown of the flagship Jev model: System One architecture versus LLMs, Choice/Score/Noul primitives, request parallelism, confidence mathematics, video use cases, code, and an overview of local open-weight alternatives (Laya, GLiNER2.5, CLM-8B).
SLMs (Small Language Models 1B–3B)
Ultra-compact next-generation models sized 1B–3B parameters (Llama 3.2, SmolLM, Qwen 2.5), designed for local execution on phones, in browsers, and on inexpensive edge servers.
ElevenLabs Updates: Comprehensive Guide to New Tools and Voice Control
Comprehensive guide to ElevenLabs speech generation: model overview (v3, Multilingual v2, Flash, Turbo), emotional audio tags, IPA/Arpabet phonemes, PLS dictionaries, and text normalization.