ResembleAI · audio tts

chatterbox

ResembleAI/chatterbox

chatterbox at Q4_K_M is exactly 491,516,160 bytes (0.46 GiB / 0.49 GB) — an effective 14.894 bits per weight, not the nominal 4.

From the file· summed from 2 file(s)
Parameters
264M
Architecture
chatterbox-s3gen
Context
—
native (config.json)
License
mit

Shipped quantizations

● exact bytes, summed from published files
QuantSize●Exact bytes●Effective bpw●Tensors●Publisher
Q2_K0.17 GiB184,593,7285.594—hans00
Q3_K_S0.20 GiB216,542,5286.562—hans00
Q3_K_M0.23 GiB242,396,4807.345—hans00
Q3_K_L0.25 GiB265,203,0088.036—hans00
Q4_00.26 GiB283,389,2488.588—hans00
Q4_K_S0.27 GiB285,486,4008.651—hans00
Q4_10.29 GiB314,846,5289.541—hans00
Q5_00.32 GiB346,303,80810.494—hans00
Q5_K_S0.32 GiB346,303,80810.494—hans00
Q5_10.35 GiB377,761,08811.447—hans00
Q6_K0.38 GiB413,150,52812.520—hans00
Q2_K2 shards0.44 GiB470,229,85614.249—calcuis
Q4_K_M2 shards0.46 GiB491,516,16014.894—hans00
Q5_K_M2 shards0.52 GiB557,936,00016.907—hans00
Q3_K_M2 shards0.54 GiB575,573,72817.442—calcuis
IQ3_XXS2 shards0.59 GiB631,606,20819.140—calcuis
IQ3_S2 shards0.63 GiB678,792,12820.569—calcuis
Q4_K_M2 shards0.66 GiB713,116,73621.610—calcuis
Q8_02 shards0.72 GiB774,652,54423.474—hans00
IQ4_XS2 shards0.73 GiB781,028,28823.667—calcuis
IQ4_NL2 shards0.76 GiB812,485,56824.621—calcuis
Q4_02 shards0.76 GiB812,485,56824.621—calcuis
Q5_K_M2 shards0.78 GiB842,437,95225.528—calcuis
Q4_12 shards0.82 GiB875,400,12826.527—calcuis
Q5_02 shards0.87 GiB938,314,68828.434—calcuis
Q6_K2 shards0.91 GiB979,841,79229.692—calcuis
Q5_12 shards0.93 GiB1,001,229,24830.340—calcuis
BF160.94 GiB1,006,916,35230.512—hans00
Q8_02 shards1.23 GiB1,315,802,048——calcuis
Q4_K4 shards1.25 GiB1,336,896,256——cstr
F162 shards1.25 GiB1,341,943,424——hans00
BF162 shards1.48 GiB1,594,485,280——calcuis
F163 shards1.49 GiB1,596,218,784——calcuis
Q8_04 shards1.87 GiB2,007,701,760——cstr
F322 shards2.44 GiB2,618,139,040——hans00
F323 shards2.97 GiB3,191,740,192——calcuis
F164 shards3.33 GiB3,575,071,616——cstr

Compare with

same modality, comparable size
vs Ace-Step1.5vs VieNeu-TTS-0.3Bvs Fun-CosyVoice3-0.5B-2512vs Kokoro-82Mvs VieNeu-TTSAll alternatives →Will it fit on my card? →

Will it run on your card?

full quant x context sweep
By VRAM:8GB12GB16GB24GB32GB48GB
Radeon RX 6500 XT 4GBGeForce RTX 3050 6GBGeForce RTX 5050 8GBGeForce RTX 3080 10GBGeForce RTX 2080 Ti 11GBGeForce RTX 5070 12GBGeForce RTX 5060 Ti 16GBApple M3 Pro 18GBGeForce RTX 3080 Ti 20GBGeForce RTX 5090 D V2 24GBGeForce RTX 5090 D 32GBApple M5 Max 36GBApple M5 Pro 48GBApple M5 Max 64GBApple M3 Ultra 96GBApple M5 Max 128GBApple M2 Ultra 192GBApple M3 Ultra 256GBApple M3 Ultra 512GB

Why other calculators give a different number

A parameters × bits ÷ 8 estimate puts Q4_K_M at roughly 0.14 GiB. The real file is 0.46 GiB, because a quantization is a mixture and some tensors are always kept at higher precision.

Architecture

Architecture unavailable — this repository is gated and no ungated mirror was found. Exact file sizes above are still authoritative; only the KV math needs the config.

Questions people ask

How much VRAM does chatterbox need?
Q4_K_M is exactly 491,516,160 bytes (0.46 GiB) in weights. Add the KV cache, which depends on your context length, plus roughly half a gigabyte of runtime overhead.
Which quantization of chatterbox should I use?
Q4_K_M is the usual default. Pick the largest quantization that fits your card at the context you actually need — the table above gives exact sizes for every one published.