If you are building a voice app today, you have a problem. Not a shortage of TTS models. The problem is too many of them, and every public leaderboard tells you a different story.
Artificial Analysis ranks models by blind preference Elo. The Humanness Index scores how human a voice sounds on a 0-100 scale. Voice Arena runs head-to-head tournaments where real people vote on which clip sounds better. All three are good at what they do. None of them will tell you which model is best across the board.
That is the gap texttospeech.com fills. We started this project because we kept flipping between tabs trying to compare the same model on different boards. It is a terrible way to make decisions.
We pull the raw numbers from every public TTS leaderboard we can find. Three so far, with more in the pipeline. Each board uses its own scoring system and covers its own set of models. Our aggregation method, Bayesian pairwise rank aggregation, converts every board’s ranking into a common 0-100 scale without losing the detail of how each model actually performed.
The details are on the methodology page. Here is what you need to know: we do not average scores. Averaging would let a single dominant placement on one board carry a model to the top. Instead we convert each board’s ranks into pairwise comparisons and compute a consensus ranking from every head-to-head match across all boards. A model that wins consistently across sources beats one that peaks on a single board and goes missing elsewhere.
Who should care? Anyone picking a TTS model for an app. If you are choosing between OpenAI, ElevenLabs, Cartesia, or SpeechifyAI right now, you can pull up the board and see how they stack up across every public benchmark. The answer will be more honest than any single board can offer.
The board refreshes daily. When a new snapshot drops, the scores recompute automatically. No manual curation, no editorial picks, no vendor bias. The same formula runs on the same data for every model, and the source snapshots are archived so you can verify any score.
Every week we will post movers and shakers. Models that gained ground, models that slipped, new entrants worth watching. The first edition covers a newcomer that hit the top spot on day one. It tells you something about where TTS is right now.
If you are comparing models for a project, this is the starting point. Pull up the board, sort by aggregate score, and dig into the source boards for the kind of quality you care about. We built this so you do not have to.