model family
MMS-TTS (Meta) logo

MMS-TTS (Meta)

Published by facebook. 12 members measured on real machines, across 12 builds. Every number below was signed by the machine that produced it — none of it is from a model card.

what it takes, what it returns
text → audio

Declared by the network’s catalog, not measured — a signature says what this family would be asked. Whether it answered is the signed capability flags on each member.

12 of 12 builds measured
what each build proved
hi
as
bn
gu
kn
ml
mr
or
pa
ta
te
ur
hi · Audio — provedhi · Speech out — provedas · Audio — provedas · Speech out — provedbn · Audio — provedbn · Speech out — provedgu · Audio — provedgu · Speech out — provedkn · Audio — provedkn · Speech out — provedml · Audio — provedml · Speech out — provedmr · Audio — provedmr · Speech out — provedor · Audio — provedor · Speech out — provedpa · Audio — provedpa · Speech out — provedta · Audio — provedta · Speech out — provedte · Audio — provedte · Speech out — provedur · Audio — provedur · Speech out — proved
Audio
Speech out
MMS-TTS Hindi1 build measured →
facebook/mms-tts-hinhi · facebook · speech-hf · 5 machines
—recalled
—served
—best tok/s
MMS-TTS Assamese1 build measured →
facebook/mms-tts-asmas · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Bengali1 build measured →
facebook/mms-tts-benbn · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Gujarati1 build measured →
facebook/mms-tts-gujgu · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Kannada1 build measured →
facebook/mms-tts-kankn · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Malayalam1 build measured →
facebook/mms-tts-malml · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Marathi1 build measured →
facebook/mms-tts-marmr · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Odia1 build measured →
facebook/mms-tts-oryor · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Punjabi1 build measured →
facebook/mms-tts-panpa · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Tamil1 build measured →
facebook/mms-tts-tamta · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Telugu1 build measured →
facebook/mms-tts-telte · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s
MMS-TTS Urdu1 build measured →
facebook/mms-tts-urd-script_arabicur · facebook · speech-hf · 1 machine
—recalled
—served
—best tok/s

Builds this network knows of

Declared, not measured — this is what the catalog says exists and what it would cost a machine to hold. Nothing here is signed, and none of it claims the build works.

how much memory does your machine have

12 of 12 builds fit a 16 GB machine — budgeting 70% of it, or 11.2 GiB, because the cache and the runtime want the rest.

  • facebook/mms-tts-asmFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-benFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-gujFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-hinFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-kanFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-malFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-marFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-oryFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-panFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-tamFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-telFP32speech-hfmeasured herepublished by facebook
    0.1 GiB
  • facebook/mms-tts-urd-script_arabicFP32speech-hfmeasured herepublished by facebook
    0.1 GiB

Sizes are what the catalog declares, not what a machine measured — fitting on disk is necessary and not sufficient. Nobody here has run these on your machine, and this narrows the candidates rather than certifying one.

the same builds, by member
MMS-TTS Assamese36M
facebook/mms-tts-asmFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Bengali36M
facebook/mms-tts-benFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Gujarati36M
facebook/mms-tts-gujFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Hindi36M
facebook/mms-tts-hinFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Kannada36M
facebook/mms-tts-kanFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Malayalam36M
facebook/mms-tts-malFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Marathi36M
facebook/mms-tts-marFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Odia36M
facebook/mms-tts-oryFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Punjabi36M
facebook/mms-tts-panFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Tamil36M
facebook/mms-tts-tamFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Telugu36M
facebook/mms-tts-telFP32speech-hf139 MiB on diskpublished by facebook
MMS-TTS Urdu36M
facebook/mms-tts-urd-script_arabicFP32speech-hf139 MiB on diskpublished by facebook
ask your own gram

Your concierge can answer this from the network rather than from this page. The ask carries the operator id, so it resolves to a capability and an engine — not to a name somebody has to recognise.

What has this network measured about the MMS-TTS (Meta) family (epn.family.mms-tts), and can my gram run any of it? See /inferences/mms-tts

open a gramx with this ask →epn.family.mms-tts