Skip to content
Local AI Wizard
Your machine
Theme

Local audio models

Speech to text · Speak, transcribe, convert audio, music and sound effects.

Documented GPU memory

Lowest published path per model · GB, log scale · lower is better · 15 of 75 models

  • Color = publisher

Source: publishers' documented requirements

Word error rate · Open ASR

0–8% · lower is better · published results, not local measurements

Source: open-asr-en-shortform · checked Oct 20, 2026

Lowest documented memory

whisper-base: 1 GB

Catalog

75 models · 15 with documented memory

Portuguese

19 models list Portuguese

Choose your machine and the charts and tables mark what fits in its memory.

Audio models

75 models · 17 reviewed · checking Sep 26, 2026 

Audio models
VibeVoice-ASR-Streaming-7BMicrosoft · MIT8.7No publisher documents a memory path for this model10 · PTSep 2026MIT
granite-speech-5.0-470m-turboctcIBM · Apache 2.00.5No publisher documents a memory path for this model1Aug 2026Apache 2.0
granite-speech-5.0-470m-turboctc-ncIBM · CC BY-NC-SA 4.00.5No publisher documents a memory path for this model1Aug 2026CC BY-NC-SA 4.0
VibeVoice-ASR-BitNetMicrosoft · MIT0.3No publisher documents a memory path for this model7 · PTJul 2026MIT
Qwen3-ASR-0.6B-hfAlibaba · Apache 2.00.8No publisher documents a memory path for this model30 · PTJun 2026Apache 2.0
Qwen3-ASR-1.7B-hfAlibaba · Apache 2.02No publisher documents a memory path for this model30 · PTJun 2026Apache 2.0
granite-speech-4.1-2bIBM · Apache 2.02.3No publisher documents a memory path for this model6 · PTApr 2026Apache 2.0
granite-speech-4.1-2b-plusIBM · Apache 2.02.1No publisher documents a memory path for this model5 · PTApr 2026Apache 2.0
VibeVoice-ASRMicrosoft · MIT8.7No publisher documents a memory path for this model51 · PTJan 2026MIT
parakeet-tdt-0.6b-v3NVIDIA · CC BY 4.00.62Neither the Hub card nor the publisher's pages name its languagesAug 2025CC BY 4.0
parakeet-tdt-0.6b-v2NVIDIA · CC BY 4.00.621May 2025CC BY 4.0
Phi-4-multimodal-instructMicrosoft · MIT5.6No publisher documents a memory path for this model23 · PTFeb 2025MIT
Whisper large-v3-turboOpenAI · MIT0.809699 · PTSep 2024MIT
Whisper large-v3OpenAI · Apache 2.01.551099 · PTNov 2023Apache 2.0
audio-flamingo-3-hfNVIDIA · Publisher terms8.3No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release datePublisher terms
Breeze-ASR-26MediaTek Research · Apache 2.01.5No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateApache 2.0
canary-1b-flashNVIDIA · CC BY 4.00.8No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateCC BY 4.0
canary-1b-v2NVIDIA · CC BY 4.01No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateCC BY 4.0
canary-qwen-2.5bNVIDIA · CC BY 4.02.6No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateCC BY 4.0
cohere-transcribe-03-2026Cohere · Apache 2.02.1No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateApache 2.0
cohere-transcribe-arabic-07-2026Cohere · Apache 2.02.1No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateApache 2.0
diar_sortformer_4spk-v1NVIDIA · CC BY-NC 4.00.1No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateCC BY-NC 4.0
diar_streaming_sortformer_4spk-v2NVIDIA · CC BY 4.00.117No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release dateCC BY 4.0
diar_streaming_sortformer_4spk-v2.1NVIDIA · Publisher terms0.117No publisher documents a memory path for this modelNeither the Hub card nor the publisher's pages name its languagesNo publisher release date recorded; the Hub repository date is not a release datePublisher terms
24 of 64
11 more repositories with limited datano parameters, documented memory or release date
Audio repositories with limited data
Confucius4-R2T2-8bitMLX Community · Publisher terms2Publisher terms
faster-whisper-large-v3SYSTRAN · MITNeither the Hub card nor the publisher's pages name its languagesMIT
faster-whisper-smallSYSTRAN · MITNeither the Hub card nor the publisher's pages name its languagesMIT
indic-conformer-600m-multilingualAI4Bharat · MITNeither the Hub card nor the publisher's pages name its languagesMIT
moonshine-tinyMoonshine · MITNeither the Hub card nor the publisher's pages name its languagesMIT
omniASR-CTC-1BMeta · Apache 2.0Neither the Hub card nor the publisher's pages name its languagesApache 2.0
omniASR-W2V-7BMeta · Apache 2.0Neither the Hub card nor the publisher's pages name its languagesApache 2.0
SenseVoiceSmallQwenAudio · Publisher termsNeither the Hub card nor the publisher's pages name its languagesPublisher terms
speaker-diarization-3.1pyannote · MITNeither the Hub card nor the publisher's pages name its languagesMIT
speaker-diarization-community-1pyannote · CC BY 4.0Neither the Hub card nor the publisher's pages name its languagesCC BY 4.0
whisper-large-v3-mlxMLX Community · MITNeither the Hub card nor the publisher's pages name its languagesMIT