Project AURIONAI Use-Cases Research & Innovation
Ampixa
speech synthesis

sanoTTS

sanoTTS is a family of tiny distilled text-to-speech voices that run without cloud or NPU on a USD 3 ESP32-S3 microcontroller or in the browser, including an Indonesian voice.

AmpixaAmpixa
Diagram of sanoTTS on a microcontroller: text in, ESP32 board, speech out.
Levelmodel
Classspeech synthesis
Parameters2.27M
Contextnot in the notes
Released13 Jul 2026
Opennessopen source
Runs onmicrocontroller, laptop, CPU server, offline capable
Statusactive

Lineage

distillsuccessor
Kokoro-82M to sanoTTS: distill, medium confidenceKokoro v0.19 to Kokoro-82M: successor, high confidenceKokoro v0.19Kokoro v0.19hexgradKokoro-82MKokoro-82Mhexgrad, Jan 2025sanoTTSsanoTTSAmpixa, Jul 2026
hexgradKokoro-82M is the parent (distill)medium confidencemodel card
Where this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.

R&D Topics

On-device text-to-speechexemplarText-to-speech for low-resource languagesproduced by

Identity

Developer
Ampixa
Family
not in the notes
Released
13 Jul 2026 sourcesregistry date
Weights published
13 Jul 2026
Record checked
11 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.

Architecture and Size

Architecture
other Distilled tiny neural TTS voices: piperlite lineage from a Piper/VITS teacher; nano lineage (mel-100, ConvNeXt1D, iSTFT) from a Kokoro teacher
Total parameters
2.27M sources
Active parameters
–
Checkpoint parameters
–
Note
Voices range from 294,279 to 2,272,145 parameters; the ESP32-S3 int8 voice has 567,008
Context length
not in the notes
Input
text
Output
audio
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).

Openness and Access

Openness
open source
Licence
GNU General Public License v3.0 sources
Access routes
open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.

Deployment

Runs on
microcontroller, laptop, CPU server, offline capable
Footprint
337 KB to 8.7 MB per voice; real time on an ESP32-S3; runs in the browser via WebAssembly; pure numpy inference
Runtimes
not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.

Indonesia

Indonesian support
listed by developer
Regional languages
not in the notes
Jakarta hosting
unknown
Languages
11 voices in English, Nepali, Hindi, Vietnamese, Indonesian and Chinese
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.

Benchmark Results

The notes record no benchmark result for this model.

Industry Use Cases

No industry use case links this model.

Sources

  1. huggingface.co/ampixa/sanoTTS https://huggingface.co/ampixa/sanoTTS
  2. huggingface.co/api/models/ampixa/sanoTTS https://huggingface.co/api/models/ampixa/sanoTTS

To cite this page: Project AURION, AI Use-Cases Research & Innovation, "sanoTTS", https://aurion.semai.id/models/ampixa-sanotts/, records of October 2026.