other
0.082B parameters
Commercial OK
Reviewed May 2026

Kokoro 82M

82M-parameter StyleTTS2-derived TTS that went viral in early 2025 for matching billion-parameter TTS quality at ~1% the size. Apache-2.0 weights, dozens of preset voice packs across English (and growing language list), and realtime synthesis on CPU.

License: apache-2.0·Context: 0 tokens
BLK · VERDICT

Our verdict

OP · Fredoline Eruo|VERIFIED MAY 29, 2026
unrated

The current default-recommend open TTS for product use. Tiny, fast, Apache-2.0, and qualitatively close to commercial APIs. Reach for XTTS or Orpheus only if you specifically need cloning or expressive control.

Overview

82M-parameter StyleTTS2-derived TTS that went viral in early 2025 for matching billion-parameter TTS quality at ~1% the size. Apache-2.0 weights, dozens of preset voice packs across English (and growing language list), and realtime synthesis on CPU.

Strengths

  • Top-tier MOS on TTS Arena leaderboards despite 82M params
  • Realtime on CPU; ~10x realtime on mid-range GPU
  • Apache-2.0, no attribution or non-commercial traps
  • Dozens of stable preset voices, no per-call cloning step

Weaknesses

  • No voice cloning — preset voices only
  • Primary language is English; multilingual packs are newer and uneven
  • Limited prosody/emotion control vs XTTS or Orpheus
  • Phonemizer dependency (espeak-ng) adds packaging friction on Windows

Quantization variants

Each quantization trades model quality for file size and VRAM. Q4_K_M is the most popular starting point.

QuantizationFile sizeVRAM required
Q4_K_M0.1 GB1 GB

Get the model

HuggingFace

Original weights

huggingface.co/hexgrad/Kokoro-82M

Source repository — direct quantization required.

Hardware that runs this

Cards with enough VRAM for at least one quantization of Kokoro 82M.

NVIDIA GB200 NVL72
13824GB · nvidia
AMD Instinct MI355X
288GB · amd
AMD Instinct MI325X
256GB · amd
AMD Instinct MI300X
192GB · amd
NVIDIA B200
192GB · nvidia
NVIDIA H100 NVL
188GB · nvidia
NVIDIA H200
141GB · nvidia
NVIDIA H200 NVL (PCIe)
141GB · nvidia
Compare alternatives

Models worth comparing

Same parameter band, plus what's one tier above and below — so you can decide what actually fits your hardware.

Step up
More capable — bigger memory footprint
Step down
Smaller — faster, runs on weaker hardware
No verdicted models in the next tier down yet.

Frequently asked

What's the minimum VRAM to run Kokoro 82M?

1GB of VRAM is enough to run Kokoro 82M at the Q4_K_M quantization (file size 0.1 GB). Higher-quality quantizations need more.

Can I use Kokoro 82M commercially?

Yes — Kokoro 82M ships under the apache-2.0, which permits commercial use. Always read the license text before deployment.

What's the context length of Kokoro 82M?

Kokoro 82M supports a context window of 0 tokens (about 0K).

Source: huggingface.co/hexgrad/Kokoro-82M

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify model claims.

Related — keep moving

Compare hardware
Buyer guides
When it doesn't work
Recommended hardware
Before you buy

Verify Kokoro 82M runs on your specific hardware before committing money.

Will it run on my hardware? →Custom hardware comparison →GPU recommender (4 questions) →