# Is Audio8-TTS-Realistic-0.6B SOTA?

Pending benchmarks. Announced Oct 29, 2026; no independent results yet.

As of Nov 3, 2026, Audio8-TTS-Realistic-0.6B is pending benchmarks on isitsota: announced Oct 29, 2026, with no independent results on the boards we track yet.

Edge0's open-weights text-to-speech model with zero-shot voice cloning from a short reference clip and its transcript, or speech without a reference. 601M parameters plus a bundled 44.1 kHz codec, 23 languages (12 more than Audio8-TTS-Preview), Apache-2.0 on Hugging Face.

Canonical: https://isitsota.com/m/edge0-audio8-tts-realistic-0-6b
Last updated: 2026-10-29
Tracking since: 2026-10-29

## Announced facts

| Fact | Value |
| --- | --- |
| Lab | Edge0 |
| Announced | Oct 29, 2026 |
| API | Not yet |
| Open weights | Published · Hugging Face (Edge0/Audio8-TTS-Realistic-0.6b), custom Transformers code (trust_remote_code) |
| Parameters | 601M (601,159,424 in the main model, excluding the audio codec) |
| Base model | Edge0/Audio8-TTS-Preview-0.6b (11 languages) |
| Languages | 23: Arabic, Cantonese, Chinese, Dutch, English, Finnish, French, German, Hindi, Indonesian, Italian, Japanese, Korean, Malay, Norwegian, Polish, Portuguese, Russian, Spanish, Swedish, Thai, Turkish, Vietnamese |
| Architecture | DualAR (inspired by Fish Audio S2 Pro): a 24-layer slow AR predicts one semantic token per frame; a 4-layer fast AR predicts 10 codec codebooks (4,096 entries each); width 896 |
| Codec | Bundled 44.1 kHz neural audio codec, about 21.5 frames per second; context up to 2,048 text/audio positions |
| Voice cloning | Zero-shot from a reference recording plus its exact transcript |
| License | Apache-2.0 |
| Serving | SGLang Omni adapter with an OpenAI-compatible speech API; free Hugging Face Space playground |
| Paper | arXiv 2610.14776, first posted Oct 10, 2026 (UTC); the weights were announced on Oct 29, 2026 |

Tags marked unsure are not stated on the lab's own pages.

## What would let us call it

Verdicts use only independent boards we already track. Two consecutive snapshots are required before any crown.

- [x] Publicly announced with a source
- [x] Usable by the public (API or weights)
- [ ] Listed on an independent board we track
- [ ] Enough results for a rank in a category
- [ ] Holds across two snapshots (crowns only)

## What the lab claims

Self-reported. Shown for context; not used in any verdict.

- Lab claim: currently #1 on the Hugging Face TTS Arena (the model card says only that it is participating and has received encouraging ratings).

## Sources

- LAB: [Samuel Zeng (Edge0 founder) announcement on X](https://x.com/SamuelZengML/status/2115881491645554870) (2026-10-29)
- LAB: [Edge0 release links on X](https://x.com/SamuelZengML/status/2115881504102551856) (2026-10-29)
- REPO: [Edge0/Audio8-TTS-Realistic-0.6b on Hugging Face](https://huggingface.co/Edge0/Audio8-TTS-Realistic-0.6b)
- REPO: [Edge0-AI/Audio8_TTS on GitHub](https://github.com/Edge0-AI/Audio8_TTS)
- LAB: [Audio8-TTS-Realistic-0.6B samples](https://edge0-ai.github.io/Audio8_TTS/realistic-0.6b/)
- PAPER: [Training a Compact, Strong, and Realistic Text-to-Speech Model (arXiv 2610.14776)](https://arxiv.org/abs/2610.14776)

Tracking continues until an independent board lists it.

## Related

- Lab: [Edge0](https://isitsota.com/labs/edge0.md)
- All pending releases: https://isitsota.com/releases.md
- JSON: https://api.isitsota.com/v1/models/edge0-audio8-tts-realistic-0-6b