whatmodel.sh / tts / Spark-TTS 0.5B
Spark-TTS 0.5B
↗ share
+ contribute
Spark-TTS 0.5B is an open-weights text-to-speech model from SparkAudio, released February 2025. Its published size is 0.5B parameters. Licence: cc-by-nc-sa-4.0.
Catalogue identity notes
SparkAudio/Spark-TTS-0.5B (Qwen2.5-0.5B-based LLM + BiCodec).
Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.
$ how it compares
The text-to-speech models ranked either side of Spark-TTS 0.5B, out of 148 with a score.
| # | model | score | where |
|---|---|---|---|
| 74 |  LEMAS-TTS (w/ prosody, multilingual_prosody) | 52±6 | local |
| 75 |  Neuphonic neu_hq | 52±3 | comparison only |
| 76 | Spark-TTS 0.5B | 52±3 | local |
| 77 |  Llasa-8B | 52±3 | local |
| 78 |  Gradium TTS | 51±3 | api $38.00/M chars |
see all text-to-speech models · find the ones that fit your machine
$ what it's good at
voice cloning▁·▁▁▁▁▁▃▇█▃▃ #29 of 37 · 81% of its score · 4 tests
Beats CosyVoice-300M here, though they rank higher overall.
| test | result | vs the others in its run | counts |
|---|---|---|---|
| Seed-TTS test-zh · speaker similarity (WavLM-large SV) +1 | 65.7% ✎ | ▁·▃··▂·█▄·▃▄ beats 2/25 | 32% |
| Seed-TTS test-en · speaker similarity (WavLM-large SV) +1 | 57.2% ✎ | ▁·▃▂▂█▆▃▅··▂ beats 1/30 | 32% |
| Seed-TTS test-en · Whisper large-v3 WER +1 | 2.3% ✎ | ▁···▂▂▂▂█▆▄▂ beats 13/35 | 10% |
| Seed-TTS test-zh · Paraformer CER +1 | 1.7% ✎ | ▁·▂▁▁▁▂▃▅█·▂ beats 13/30 | 7.1% |
other tests2 not in a task · 14% of its score
| test | result | vs the others in its run | counts |
|---|---|---|---|
| RVCBench LibriTTS clean voice cloning · ECAPA speaker similarity | 41.0% ✎ | ▃·····▃▂▂▃█▆ beats 2/15 | 14% |
| RVCBench LibriTTS clean voice cloning · Whisper-medium WER | 33.0% ✎ | ▃▂▂··▂·▃▃▃█▂ beats 0/15 | 0.2% |
$ where to run it
Other ways to run it
version
- Run it as published · needs ~3.92 GB · from github.com/SparkAudio (official)
git clone https://github.com/SparkAudio/Spark-TTS.git && cd Spark-TTS && conda create -n sparktts -y python=3.12 && conda activate sparktts && pip install -r requirements.txt && python -c "from huggingface_hub import snapshot_download; snapshot_download('SparkAudio/Spark-TTS-0.5B', local_dir='pretrained_models/Spark-TTS-0.5B')" && cd example && bash infer.sh
Official weights and documentation: huggingface.co/SparkAudio ↗
$ how the score was worked out · DMOSpeech 2: Reinforcement Le… 81%, RVCBench: Benchmarking Robust… 14%
Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works
| source | results | share of the score |
|---|---|---|
 DMOSpeech 2: Reinforcement Learning for Duration Prediction in Metric-Optimized Speech Synthesis (arXiv:2507.14988v1) | 4 | 81% |
| RVCBench: Benchmarking Robustness of Voice Cloning Across Modern Audio Generation Models (arXiv:2602.00443v2) | 2 | 14% |
| prior · pull toward the middle | 4% | |
| weighted average | = 51.7 |
not counted (4)
| Spark-TTS: An Efficient LLM-Based Text-to-Speech Model with Single-Stream Decoupled Speech Tokens (arXiv:2503.01710v1) | Seed-TTS test-zh · Paraformer CER [correct] | 1.2% | the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts |
| Spark-TTS: An Efficient LLM-Based Text-to-Speech Model with Single-Stream Decoupled Speech Tokens (arXiv:2503.01710v1) | Seed-TTS test-zh · speaker similarity (WavLM-large SV) [correct] | 67.2% | the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts |
| Spark-TTS: An Efficient LLM-Based Text-to-Speech Model with Single-Stream Decoupled Speech Tokens (arXiv:2503.01710v1) | Seed-TTS test-en · Whisper large-v3 WER [correct] | 2.0% | the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts |
| Spark-TTS: An Efficient LLM-Based Text-to-Speech Model with Single-Stream Decoupled Speech Tokens (arXiv:2503.01710v1) | Seed-TTS test-en · speaker similarity (WavLM-large SV) [correct] | 58.4% | the only model in this source's run of the test, so there is nothing to compare it with there; another run of the test counts |
$ improve this page
Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.
data updated · sources and licences · json