sanoTTS

सानो — small

a tiny neural text-to-speech voice · sano (सानो) is Nepali for small

The smallest neural text-to-speech family we know of — and it plays right here, live, in your browser.

a full neural stack — duration → acoustic → decoder — quantized to int8, phonemizer and all. 28 voices, 16 languages, all spoken on this page.

Synthesize live in your browser

Type anything. The text is phonemized by espeak-ng compiled to WebAssembly (the same 275 KB port that runs on the chip; Chinese uses a 448 KB pinyin lexicon instead) and spoken by that voice's own neural stack — all client-side, no server. Pick a voice:

loading runtime…

Phonemes	—
Audio produced	—
Time (this CPU)	—
Speed on this machine	—

The speed shown is this browser on your CPU — the on-device English voice runs at 0.22× real time on an ESP32-S3. Every voice in the picker synthesizes your typed text live, right here, with no server involved. Where a language offers two sizes, the chips under its mascot switch between them. heart and heart-nano take a different route from text to sound than the other voices (a 100-band mel interface into a noise-shaping decoder), so they load their own small runtime; heart-nano runs the int8 arithmetic of the microcontroller build unchanged.

smallest neural TTS family known — 294k to 2.3M parameters
runs real-time on a $3 microcontroller (ESP32-S3)
runs right in your browser — WebAssembly, no server
under 4 MB per voice, zero dependencies (espeak-ng phonemizer included)
11 voices across 6 languages — English, Nepali (नेपाली), Hindi (हिन्दी), Vietnamese (Tiếng Việt), Indonesian (Bahasa), Chinese (中文) — all synthesizing live on this page
new: ten more languages — German, French, Spanish, Italian, Portuguese, Russian, Czech, Romanian, Turkish and Arabic — added 2026-09-08, each in a 1.56M and (for eight of them) a ~511k size, all synthesizing live in the picker above. That is 16 languages in the browser. What they score, and what that score does not tell you
new: heart, our best-sounding voice at 2.27M parameters and 24 kHz, and heart-nano, the same voice squeezed into 294k parameters — a complete text-to-speech stack, int8, in 337 KB. Both synthesize live in the picker above
new, experimental: trellis — a whole new architecture that fits a complete voice in under 1M parameters and runs about 5× faster in the browser than our 1.4M–1.8M voices. It synthesizes live up there too — a research preview, not yet as natural-sounding as amy
open source, GPL-3.0

How it compares

Measured with our no-reference harness (SCOREQ / UTMOS / DNSMOS-SIG) on the same 24-sentence diverse set, run identically across every model below.

Size comparison (parameters)

Kokoro is 36× larger than our largest voice — and 279× larger than our smallest.

Shipped-file sizes: sanoTTS amy 2.8 MB fp16 and TinyTTS 3.5 MB fp16, both verified from the released files; Kokoro's ~330 MB fp32 is its widely cited public figure.

Three metrics, head to head

	sanoTTS amy1.46M params	sanoTTS heart2.27M params	TinyTTS1.62M params	Inflect Nano4.63M params
SCOREQ	4.13	3.48	3.94	3.81
UTMOS	4.10	3.38	3.65	3.65
DNSMOS-SIG	3.61	3.51	3.62	3.58

Same 24-sentence set, three independent no-reference metrics; best value per row in crimson. sanoTTS amy leads on SCOREQ and UTMOS; on DNSMOS-SIG TinyTTS edges us by 0.01 — no single metric tells the whole story. heart is a different voice and recipe (24 kHz, noise-shaping decoder) scored on the identical 24 sentences; heart-nano, its 294k sibling, scores 2.29 / 2.45 / 3.35 on the same three metrics.

The size-quality frontier we traced

Same voice (amy), same duration/acoustic recipe — only decoder size changes. Quality lives in the decoder: doubling it from 1.09M to 1.84M params moves SCOREQ from 3.70 to 4.16.

Every language we ship, and what we know about it

WER is Whisper word error rate over 16 held-out Tatoeba sentences per language — conversational text, deliberately a different domain from the corpus these voices were distilled on. The sample buttons synthesize that language on this page, right now, through the same runtime as the picker above. Where a language ships in two sizes, both are playable. Chinese is the one exception: written Chinese has no spaces, so word error rate means nothing there and the figure is character error rate over the same 16 Tatoeba sentences, with traditional and simplified folded together. On held-out FLORES that voice scores 0.199 against its teacher’s 0.188 — the voice it replaced scored 0.468.

Language	WER	Sample
Portuguese 🇧🇷	0.038
Italian 🇮🇹	0.080
German 🇩🇪	0.099512k · 0.145
Russian 🇷🇺	0.120
Spanish 🇪🇸	0.147
French 🇫🇷	0.221
Indonesian 🇮🇩	0.256
Arabic 🇯🇴	0.274
Turkish 🇹🇷	0.303
Czech 🇨🇿	0.351
Romanian 🇷🇴	0.392
Vietnamese 🇻🇳	0.468
English 🇺🇸	—
Nepali 🇳🇵	—
Hindi 🇮🇳	—
Chinese 🇨🇳	0.024CER, not WER

Lower is better. Press play to judge the rest.

What these numbers are, and are not. WER is measured with Whisper over 16 held-out Tatoeba sentences per language — directional, not precise. It says the words come through; it says nothing about whether a voice sounds good. No listening test has been run on any of the ten languages added on 2026-09-08, and none of them has a SCOREQ. Czech, Romanian and French got worse on unseen text than on the corpus they were trained on, so they are the ones to press first. Indonesian and Vietnamese are the espeak arm of a paired A/B — the phoneme path this page actually runs. Everything above is reproducible from experiments/evidence/ood-tatoeba-20260908.json, idvi-espeak-free-ab-20260904.json and heart-diverse24-remeasure-20260904.json.

Open source under GPL-3.0 — builds on piper and espeak-ng. Distill your own tiny voice: github.com/Ampixa/sanoTTS.