FreyaTTS Technical Report
Quick Answer
Freya-TTS is a 183.2M-parameter Turkish-first text-to-speech model that achieves a WER of 8.0% and CER of 3.0% on the Freya-TR-Eval benchmark, outperforming larger systems.
Quick Take
It operates efficiently on consumer GPUs with a real-time factor of 0.11, making it suitable for edge deployment. The model and code are available under the Apache-2.0 license.
Key Points
- Freya-TTS uses a non-autoregressive conditional flow-matching Diffusion Transformer.
- The model operates in a frozen continuous latent space of AudioVAE2.
- It features rule-free end-to-end modeling with a 92-symbol Turkish character vocabulary.
- Achieves a real-time factor of 0.11 on consumer GPUs, suitable for edge devices.
- Model weights and code are released under the Apache-2.0 license.
Paper Resources
📖 Reader Mode
~2 min readAbstract:We introduce Freya-TTS, a compact, tokenizer-free, Turkish-first text-to-speech model designed for highly reliable and efficient conversational synthesis. Freya-TTS is a 183.2M-parameter non-autoregressive conditional flow-matching Diffusion Transformer (DiT) that operates in the frozen continuous latent space of AudioVAE2 (16 kHz encode, 48 kHz decode), allowing the model to focus its capacity on text-to-latent mapping while inheriting high-quality 48 kHz reconstruction. We advance the framework along three key dimensions: (1) rule-free end-to-end modeling from a 92-symbol Turkish character vocabulary without a phonemizer, grapheme-to-phoneme frontend, or discrete speech tokenizer; (2) non-autoregressive parallel denoising, which predicts the entire latent sequence simultaneously over a predicted duration; and (3) a production-oriented two-stage post-training recipe consisting of single-speaker voice locking and short-utterance coverage, improving speaker consistency and robustness on short inputs. On the Freya-TR-Eval benchmark, Freya-TTS achieves a band-matched word error rate (WER) of 8.0% and character error rate (CER) of 3.0%, outperforming substantially larger open-source systems while using a fraction of their parameters. The model achieves a real-time factor of 0.11 on consumer GPUs and runs faster than real time on a laptop CPU, making it well suited for resource-constrained edge deployment. We release the model weights, training and inference code, and evaluation benchmark under the Apache-2.0 license.
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2607.09530 [cs.CL] |
| (or arXiv:2607.09530v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.09530 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Ahmet Erdem Pamuk [view email]
[v1]
Fri, 10 Jul 2026 15:36:30 UTC (351 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.