Workflows/Transcribe audio locally with Whisper large-v3-turbo
Audio / OpenAI
Transcribe audio locally with Whisper large-v3-turbo
Import a recording into Buzz, run Whisper large-v3-turbo locally, review the transcript and export TXT and SRT without building a Python environment.
EstimatedSource checked 10/14/2026Version: Reviewed easiest local route · 2026-10-14
LOCAL
Freshness 6 out of 100. Repository age: 744 days old, Oct 1, 2024. Measures age, not quality.6/ 100Freshness744 days oldRepository age · Oct 1, 2024
Popularity 92 out of 100. 6,349,781 downloads in 30 days; 3,454 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.92/ 100Popularity6.3M downloads / 30d3.5K lifetime likes · Checked Oct 15, 2026
What do these scores mean?+
Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.
Freshness · release age
A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.
Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.
Scores as of Oct 15, 2026. Statistics checked Oct 15, 2026. Statistics are marked for refresh after 72 hours.Full scoring method →
Before you begin
Required VRAM
8 GB
Hardware
The model is the same across all three desktop platforms. Choose the backend available in Buzz on your machine; these are starting memory allowances, not speed guarantees.
Machine
Whisper large-v3-turbo · Buzz desktop
NVIDIA GPU · 6–8 GB VRAM + 16 GB RAM
Local CUDA-capable route; 6 GB is the publisher tier, 8 GB provides margin.
AMD or Intel GPU PC · 16 GB RAM
Use a compatible local Buzz backend; CPU remains a fallback. GPU acceleration depends on the chosen backend.
Apple Silicon Mac · 8–16 GB unified memory
Local desktop route; allow 16 GB for more headroom.
CPU-only PC · 8–16 GB RAM
Transcription is possible but may take longer; start with a short file.
The downloadable weight size does not include runtime and audio memory. Review the first completed result before submitting long recordings.
Tradeoffs
Transcribes speech; the turbo checkpoint is not the translation model.
Silence or noisy speech can produce invented words; review the transcript and subtitle timing.
This route does not separate speakers or provide a live microphone transcription service.
LOCAL AUDIO TRANSCRIPTIONSOURCE-LINKED SETUP
Transcribe audio locally with Whisper large-v3-turbo
Install Buzz once, import a speech recording, select the exact local Whisper large-v3-turbo model, review the result and export TXT plus SRT. No Python, FFmpeg configuration or terminal commands are needed for the normal route.
Choose a package Use the built-in workspace Check the result
01
BEFORE YOU BEGIN
The documented package
This setup uses the package documented for this task. Review its source and supported platforms before starting.
Save your machine in My Hardware to get an automatic starting choice. You can always choose any package yourself.
OpenAI lists about 6 GB VRAM for large-v3-turbo; CPU and Apple Silicon can also transcribe with enough host memory. Actual speed depends on Buzz backend, device and file length.
Freshness 6 out of 100. Repository age: 744 days old, Oct 1, 2024. Measures age, not quality.6/ 100Freshness744 days oldRepository age · Oct 1, 2024
Popularity 92 out of 100. 6,349,781 downloads in 30 days; 3,454 lifetime likes. Checked Oct 15, 2026. Measures adoption, not quality.92/ 100Popularity6.3M downloads / 30d3.5K lifetime likes · Checked Oct 15, 2026
What do these scores mean?+
Two ways to compare models. Neither score measures answer quality, benchmark performance, or whether a model fits your machine.
Freshness · release age
A new release starts at 100. The score halves every 180 days: 100 × 2^(−age in days / 180). When the release date is unavailable, a repository date is a fallback—not proof of when the model was released. Missing dates stay unscored.
Downloads over the past 30 days carry 80% of the score; lifetime likes carry 20%. Both use a logarithmic scale, capped at 10 million downloads and 100,000 likes. If likes are unavailable, downloads carry all the weight. Missing download statistics stay unscored.
Set up Whisper large-v3-turbo · Buzz desktop on your machine
Choose where you want to run it. Use the built-in workspace; no workflow import is needed.
CHOOSE YOUR SETUPWindows
01
Install the Buzz desktop app
Download the Windows ZIP Buzz 1.4.5 build linked by the Buzz publisher, extract the archive and open the app. Intel Mac owners should choose the separate macOS-X64 archive. The app runs transcription on your machine; choose a local Whisper backend rather than the OpenAI API. The first model download requires internet, but file transcription works offline afterward.
Import a short recording and choose the exact model
In Buzz, import a clear 10–30 second speech file. Set the task to Transcribe, select the local OpenAI Whisper backend and large-v3-turbo model, and choose the source language. Buzz lists this checkpoint explicitly. Confirm the model download finishes before starting the job; avoid the API backend if you want local processing.
Click Transcribe in the Buzz window to process the selected recording with large-v3-turbo. Open the result in its transcription viewer, play the source recording and check several words and timestamps against what you hear. Correct mistakes before using the text; noisy audio can produce hallucinated words.
Use Buzz’s Export action to save both input.txt and input.srt in a folder you choose. Open input.txt and play the audio at a few SRT cue times to confirm the exported words and timing. Buzz also supports VTT if your editor prefers it. Use your recording’s actual filename stem when the export dialog offers a default.
Buzz exports TXT and SRT for the same recording. The words match what you hear, and several subtitle cues line up when played against the source audio.
Buzz is an independently maintained desktop interface around the OpenAI Whisper checkpoint. The original publisher command-line route remains a separate advanced option, not the first-run path.
The model selector shows only an API route or a different Whisper size.
Choose the local OpenAI Whisper backend and large-v3-turbo. Verify the exact model appears in the current Buzz build; do not substitute large-v3 or an API model.
The model download or file import fails.
Check internet access for the first model download, then import a short playable WAV or MP3. Reopen Buzz after the model cache completes.
The transcript contains invented words or subtitle timing is wrong.
Replay the source in the Buzz viewer, correct the text and re-export TXT and SRT. Shorten long silence or improve the recording before rerunning.
REFERENCE LIBRARY
Sources and files
Original instructions, model files, and compatibility notes behind this setup.