The best text-to-speech models, according to hackers

18 text-to-speech models recommended in Hacker News comments over the past year, ranked by how many separate people recommend each one.

All text-to-speech models

ElevenLabsPeople recommend it as the top-quality text-to-speech and voice cloning service, with natural voices and easy custom voices. The most common caveat is per-character pricing that gets expensive at high volume.59recommend
Kokoro TTSPeople recommend Kokoro for natural-sounding speech that runs locally on a CPU, is small, and is free and open-source. Common caveat: mistakes on some common phrases and numbers.59recommend
Pocket TTSSmall, fast local text-to-speech model that runs on CPU and clones voices from short clips. Critics note skipped words and a prohibited-use clause.41recommend
Qwen3-TTSOpen-source text-to-speech model that clones voices well and runs locally, though it needs a GPU for real-time use and can be inconsistent across clips.41recommend
Kitten TTSOpen-source text-to-speech models under 25MB that people find impressively good for their size. Common caveats: cartoonish voices and trouble with numbers and domain-specific words.33recommend
Piper TTSFast, free, open-source neural text-to-speech with usable out-of-box voices, mostly for notifications and system messages. Common caveat: voices sound robotic or machine-like.21recommend
Chatterbox TTSPeople recommend Chatterbox for high-quality, natural-sounding speech and voice cloning that runs locally. The most common caveat is that it is slower than some alternatives.13recommend
Sopro TTSSmall open text-to-speech model with voice cloning, free and easy to run locally; people like its size and speed, but most critics find the audio quality poor and robotic.13recommend
Microsoft VibeVoiceFree, MIT-licensed Microsoft speech model praised for expressive output and built-in diarization; critics report random music artifacts and thin documentation.10recommend
Supertonic TTSPeople recommend it as a very fast, tiny open-source text-to-speech model that runs well on CPU. Critics say its audio quality is only decent for its speed and it lacks extra features.9recommend
Inflect-Micro-v2Free open-weight speech model that people find surprisingly natural for its small size. Some find the inflections odd and not very enjoyable to listen to.9recommend
Gemini 3.8 text-to-speechPeople praise its expressive, natural-sounding speech and its large voice library with tight control. Common caveat: it is not production ready and can sound artificial.8recommend
ElevenReaderRecommended for turning web novels, essays and text into narrated audio, with a useful free tier. Critics say its voices trail NotebookLM and it lacks interactivity.7recommend
Fish AudioPeople recommend it for natural-sounding speech, especially long narration, and fine-grained control of tone and emotion. No common caveat.7recommend
SpeechifyPeople recommend it for good voice quality at a reasonable price, with several saying it looks promising. Some say cheaper or better alternatives may exist.7recommend
OmniVoiceText-to-speech model people recommend for fast, high-quality speech, with voice cloning and self-hosted use.5recommend
Paper2AudioA text-to-speech service built to read complex documents such as research papers aloud, with a browser extension that adds open tabs to a listening queue.5recommend
15.aiPeople recommend 15.ai as a landmark of projects built for their own sake, a free voice tool that was popular in its time. One commenter says its creator was well placed but missed the opportunity.5recommend