# Seed-TTS
Seed-TTS is a family of large-scale speech generation models developed by ByteDance's Seed team, first described in a June 2024 technical report, that generates natural speech in a zero-shot way: given only a few seconds of reference audio, it clones a speaker's voice without any speaker-specific fine-tuning. The family sits within a broader Seed audio portfolio that also includes speech recognition (Seed-ASR) and music generation (Seed-Music) models; Seed-ASR capabilities have been commercialized in China through Doubao speech recognition models, while Seed-TTS provides speech generation.[1](https://arxiv.org/html/2406.02430v1) • [2](https://arxiv.org/pdf/2407.04675v2)
Seed-TTS is trained on data volumes the report describes as "orders of magnitudes larger than the previously largest TTS systems", which the authors credit for strong generalization and emergent abilities.[1](https://arxiv.org/html/2406.02430v1)
| Fact | Detail |
| --- | --- |
| First release | Seed-TTS technical report, June 2024[1](https://arxiv.org/html/2406.02430v1) |
| Architecture | Autoregressive transformer with tokenizer, token LM, token diffusion model, and vocoder; a diffusion-only Seed-TTS DiT variant also exists[1](https://arxiv.org/html/2406.02430v1) |
| Zero-shot cloning | Reference utterances of 3 to 20 seconds[1](https://arxiv.org/html/2406.02430v1) |
| Seed-ASR training data | Over 20 million hours of speech, nearly 900 thousand hours of paired ASR data[2](https://arxiv.org/pdf/2407.04675v2) |
| Availability | Weights and code not released, citing AI safety; commercial access via BytePlus Seed Speech[3](https://whisper-ai.app/blog/seed-audio-1-0-bytedance-beyond-tts) |
| Benchmark status | All published benchmark numbers are vendor-reported; no independent evaluation appears in the sources reviewed[1](https://arxiv.org/html/2406.02430v1) • [2](https://arxiv.org/pdf/2407.04675v2) |
## Releases and versions
The family followed a steady cadence through 2024 and 2025:
- **Seed-TTS (June 2024)**: the founding speech generation report, introducing the autoregressive family and the Seed-TTS DiT variant.[1](https://arxiv.org/html/2406.02430v1)
- **Seed-ASR (July 2024)**: LLM-based speech recognition in Chinese multi-dialect and multilingual variants.[2](https://arxiv.org/pdf/2407.04675v2)
- **Seed-Music (September 2024)**: a suite of music generation systems combining autoregressive language modeling and diffusion, with fine-grained style control and post-production editing of lyrics and vocal melodies.[4](https://arxiv.org/html/2409.09214v1)
- **Seed Audio 1.0 (2025)**: an audio creation model evaluated by [ByteDance](https://www.edgepedia.org/bytedance) across text-prompted voice generation, multi-scenario audio creation, and multilingual generation.[5](https://seed.bytedance.com/en/blog/from-speech-to-audio-creation-introducing-the-seed-audio-1-0-audio-creation-model)
- **Seed-LiveInterpret 2.0**: an end-to-end simultaneous interpretation model with speech-to-speech generation and voice cloning, described by ByteDance as a fully operational product-level solution.[6](https://seed.bytedance.com/en/direction/speech)
The sources reviewed do not document Seed-TTS-0.5, Seed-TTS 1.0, or a Seed-TTS 2.0 release, so any such versions cannot be described here.
## How it works
Seed-TTS is an autoregressive transformer system built from four blocks: a speech tokenizer that converts audio into discrete tokens, a token language model that predicts those tokens from text, a token diffusion model, and an acoustic vocoder that renders the waveform.[1](https://arxiv.org/html/2406.02430v1) It is therefore a hybrid: an LLM-style token stage followed by diffusion-based refinement, in the same general pattern as the Seed-Music pipeline, whose four blocks (audio tokenizer, autoregressive LM generator, token diffusion model, vocoder) render 44.1 kHz stereo audio.[4](https://arxiv.org/html/2409.09214v1)
Training proceeded in three stages: pre-training, fine-tuning (speaker and instruction fine-tuning), and post-training via reinforcement learning.[1](https://arxiv.org/html/2406.02430v1)
**Zero-shot voice cloning** works from reference utterances of 3 to 20 seconds; no per-speaker training is required.[1](https://arxiv.org/html/2406.02430v1) The later Seed Audio 1.0 extends the zero-shot idea to voice description: creators can shape a target voice through text alone or pair a description with a reference sample, again with no separate model training per voice (vendor-reported).[5](https://seed.bytedance.com/en/blog/from-speech-to-audio-creation-introducing-the-seed-audio-1-0-audio-creation-model)
A non-autoregressive variant, Seed-TTS DiT, uses a fully diffusion-based architecture and does not depend on pre-estimated phonemes, offering an alternative generation path within the same family.[1](https://arxiv.org/html/2406.02430v1)
Seed-ASR takes a different architectural route: an audio-conditioned LLM (AcLLM) framework that feeds continuous speech representations plus contextual information into the LLM, rather than operating on discrete audio tokens. Its training recipe runs from self-supervised learning of the audio encoder through supervised fine-tuning, context SFT, and reinforcement learning.[2](https://arxiv.org/pdf/2407.04675v2)
## Benchmarks: vendor-reported versus independent
Every quantitative claim below is vendor-reported. No independent evaluation of Seed-TTS or Seed-ASR appears in the sources reviewed, which is the single most important caveat for reading these numbers.
**Seed-TTS evaluation methodology.** ByteDance measured word error rate (WER) using Whisper-large-v3 for English and Paraformer-zh for Mandarin, speaker similarity using WavLM-large embeddings, and human preference via comparative mean opinion score (CMOS) studies on a -2 to +2 scale, where an absolute CMOS below 0.1 is considered insignificant between two systems.[1](https://arxiv.org/html/2406.02430v1)
**Seed-TTS-Eval.** ByteDance released the configuration of the objective test set in a GitHub repository to enable benchmarking by others, but withheld the subjective set due to copyright restrictions.[1](https://arxiv.org/html/2406.02430v1) The sources reviewed do not describe the test set's composition or document its adoption as a standard benchmark, so its status as one cannot be confirmed here.
**Seed-ASR comparisons.** ByteDance reports that Seed-ASR achieves a 10% to 40% reduction in word error rates (character error rates for Chinese) versus recently released large ASR models on Chinese and English public test sets.[2](https://arxiv.org/pdf/2407.04675v2) In the paper's comparison table, on English multi-domain WER, Seed-ASR (ML) scored 5.34% against 10.41% for Whisper Large v3 and 9.33% for Google USM.[2](https://arxiv.org/pdf/2407.04675v2) These figures come from ByteDance's own evaluation, not from independent replication.
**Seed Audio 1.0.** ByteDance reports that on audio naturalness, most languages achieved MOS scores above 4.0.[5](https://seed.bytedance.com/en/blog/from-speech-to-audio-creation-introducing-the-seed-audio-1-0-audio-creation-model)
## Availability and adoption
Seed-TTS is closed-weights: the public repository notes that source code and model weights were not released due to AI safety considerations, though the objective test-set configuration and metric scripts (WER, speaker similarity) were published for evaluation.[3](https://whisper-ai.app/blog/seed-audio-1-0-bytedance-beyond-tts) • [1](https://arxiv.org/html/2406.02430v1) This places it apart from open-weights speech models; the sources reviewed provide no pricing information and no head-to-head third-party comparison with [ElevenLabs](https://www.edgepedia.org/elevenlabs), OpenAI TTS, or [CosyVoice](https://www.edgepedia.org/cosyvoice).
Commercial access runs through BytePlus, ByteDance's global cloud service, which presents Seed Speech as a product family with text-to-speech, voice replication, and speech-to-text capabilities, with SeedTTS positioned for human-like speech and voice replication and SeedASR for recognition.[3](https://whisper-ai.app/blog/seed-audio-1-0-bytedance-beyond-tts) In China, Seed-ASR capabilities have been applied across a variety of ByteDance products and commercialized through Doubao speech recognition models.[2](https://arxiv.org/pdf/2407.04675v2) No quantitative usage figures for Doubao or other ByteDance products appear in the sources reviewed.
## References
1. [Seed-TTS: A Family of High-Quality Versatile Speech Generation Models](https://arxiv.org/html/2406.02430v1)
2. [Seed-ASR: Understanding Context and Supporting Diverse Languages and Accents](https://arxiv.org/pdf/2407.04675v2)
3. [Seed Audio 1.0: How ByteDance Is Moving Beyond Text-to-Speech](https://whisper-ai.app/blog/seed-audio-1-0-bytedance-beyond-tts)
4. [Seed-Music: A Unified Framework for High Quality and Controlled Music Generation](https://arxiv.org/html/2409.09214v1)
5. [From Speech to Audio Creation | Introducing the Seed Audio 1.0 Audio Creation Model](https://seed.bytedance.com/en/blog/from-speech-to-audio-creation-introducing-the-seed-audio-1-0-audio-creation-model)
6. [ByteDance Seed — Speech direction](https://seed.bytedance.com/en/direction/speech)
---
*Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › Model families and named models › Audio, music, and speech models*
*Initially written Sep 17, 2026 · Reviewed: Sep 30, 2026 · Edited: Sep 30, 2026; Oct 11, 2026; Oct 25, 2026 · Last review: Sep 30, 2026*
*Copyright 2026 EdgeChat AI, a subsidiary of Biostate AI.*
License: Edgepedia Community License 1.0, https://www.edgepedia.org/edgepedia/license