Comment transcrire automatiquement des enregistrements Audacity en 2026

One of the best ways to transcribe Audacity recordings automatically is Sonix, which returns speaker-labeled, time-stamped transcripts with jusqu'à une précision de 99% and can process a one-hour file in under five minutes. For free, offline transcription without uploading files anywhere, Audacity’s built-in OpenVINO Whisper plugin is the leading Audacity-native option that runs entirely on your own computer.

There are two proven paths for how to transcribe Audacity recordings automatically in 2026. The first runs entirely on your computer using Audacity’s free OpenVINO Whisper plugin: no uploads, no subscription, no account required. The second exports your file and uploads it to Sonix, which returns a highly accurate, speaker-labeled, time-stamped transcript in under five minutes with export options for DOCX, PDF, SRT, VTT, and 30+ other formats. Sonix accepts Audacity’s native AU format directly, so no conversion step is required.

This guide covers both methods end-to-end: prerequisites, step-by-step instructions, and a verdict on which approach fits your workflow. Based on our evaluation of both methods, Sonix is the stronger choice for anyone who needs multi-speaker support, professional export formats, or up to 99% accuracy on clean audio.

En bref : Export your Audacity recording as MP3 or WAV, upload it to Sonix, select the language, and click Start Transcribing. A 60-minute file returns a speaker-labeled, time-stamped transcript in under five minutes (up to 10x faster than real time, depending on conditions). For offline transcription, use Audacity’s free OpenVINO Whisper plugin (Analyze > OpenVINO Whisper Transcription). Full step-by-step instructions for both methods below.

Principaux enseignements

• Audacity has a free built-in transcription option via the OpenVINO Whisper plugin: it runs entirely on your own computer with no file uploads
• The OpenVINO plugin requires a several GB download (size varies by model selection: base model is under 1 GB; full install with all models can exceed 6 GB), manual setup, and produces a raw label track rather than formatted, paragraph-ready text
• Sonix accepts Audacity’s native AU format directly, so you can skip the export step entirely
• Sonix processes a 60-minute recording up to 10x faster (a one-hour file can be ready in under five minutes, depending on conditions) with up to 99% accuracy on clean audio across 53+ langues
• AI speaker diarization automatically labels each speaker’s words in the transcript: no manual tagging required
• Sonix exports to 30+ formats, including DOCX, PDF, SRT, VTT, and plain text: ready for YouTube, Premiere, or legal documentation
• Trusted by Google, Stanford, and ESPN, with 6.2M+ users who have transcribed more than 14.2M+ hours of audio (Sonix-reported)

Quick Steps to Transcribe Audacity Automatically

Method 1: Free and Offline (Audacity OpenVINO Whisper Plugin)

1. Download and install the OpenVINO AI Plugin from audacityteam.org
2. Open your Audacity recording and select the audio (Ctrl+A)
3. Aller à Analyze > OpenVINO Whisper Transcription
4. Choose a Whisper model size and set the language manually
5. Cliquez sur Apply: the transcript appears as a label track beneath the audio
6. Export via File > Export Other > Export Labels (SRT, VTT, or plain text)

Method 2: Cloud, 99% Accurate (Sonix)

1. Export your recording from Audacity as MP3 or WAV, or skip this step entirely since Sonix accepts .AU directly
2. Upload the file to sonix.ai
3. Select the recording language from 53+ supported options
4. Cliquez sur Start Transcribing: a 60-minute file returns a labeled transcript in under five minutes
5. Review and edit in Sonix’s synchronized browser editor
6. Export as DOCX, PDF, SRT, VTT, or 30+ formats

What You Need Before Starting

For Method 1 (Audacity’s Built-In OpenVINO Whisper Plugin)

• A current Audacity version installed on your Windows, macOS, or Linux computer (use the version recommended by the OpenVINO plugin release notes; current Audacity releases are in the 3.7.x line)
• The OpenVINO AI Plugin for Audacity, downloaded from the official Audacity plugins page at audacityteam.org
• A recorded audio track open in Audacity
• Sufficient processing power: the medium and large Whisper models require more CPU/GPU resources

For Method 2 (Export and Upload to Sonix)

• A recorded audio file in Audacity (or an already-exported MP3, WAV, M4A, AU, FLAC, OGG, or AAC file)
• A Compte Sonix: the free trial includes 30 minutes of transcription with no credit card required
• Knowledge of the language(s) spoken in the recording

Does Audacity Have Built-In Transcription?

Yes. Since the introduction of the OpenVINO AI plugin, Audacity supports on-device transcription powered by OpenAI’s Whisper model. The plugin installs directly into Audacity and adds a new analysis effect called OpenVINO Whisper Transcription under the Analyze menu. Transcription runs entirely on your own computer: no internet connection is needed, no audio leaves your machine, and there are no per-minute charges.

The built-in method writes the transcript to a label track below the audio waveform. Each phrase appears as a time-stamped label that you can edit and export as SRT, VTT, or plain text using File > Export Other > Export Labels.

However, most professionals who need Audacity audio transcription at scale, or who need speaker diarization, 53+ language support, a synchronized editing interface, or export to Word and PDF, use a dedicated automated transcription service alongside Audacity rather than the plugin alone.

Method 1: Free, Offline Transcription (OpenVINO Whisper)

This method runs entirely on your computer. No account, no uploads, no subscription. Best for single speakers, clean audio, and offline environments.

Step 1: Install the OpenVINO AI Plugin for Audacity

1. Go to the OpenVINO AI Effects for Audacity page on audacityteam.org
2. Download the installer for your operating system (Windows, macOS, and Linux are all supported)
3. Run the installer: it places the plugin files in Audacity’s plugins directory automatically
4. Launch or restart Audacity
5. Open the Analyze menu: you should now see OpenVINO Whisper Transcription and related AI effects listed

If the menu item doesn’t appear, go to Tools > Plugin Manager, locate the OpenVINO entries, click Enable, and restart Audacity.

Step 2: Reduce Noise Before Transcribing

Whisper performs best on clean audio. Before running transcription, denoise the track:

1. Select all audio on the track (Ctrl+A or Cmd+A)
2. Aller à Effect > OpenVINO Audio Denoising (installed with the plugin pack)
3. Apply the effect and let it process
4. Play back a few seconds to confirm the background noise has been reduced

Skipping this step on recordings with significant background noise (fan hum, room reverb, outdoor wind) measurably reduces transcription accuracy.

Step 3: Select Your Audio and Run Whisper Transcription

1. Select the audio region you want to transcribe: use Ctrl+A to select the entire track, or drag to select a specific segment
2. Aller à Analyze > OpenVINO Whisper Transcription
3. In the dialog that opens, choose your settings:
   • Model size: base is fastest for clean English; small or medium handles accents and non-English languages better; large is the most accurate but slowest
   • Language: Set this manually, especially for short clips, to prevent Whisper from misdetecting the language
   • Initial Prompt (optional): Enter proper names, brand names, or technical terms so Whisper spells them correctly throughout the transcript
   • Identification des locuteurs : Enable this if you need speakers labeled (requires the small.en-tdrz model, which includes experimental diarization)
4. Cliquez sur Apply and wait while the model processes the audio

When transcription completes, a new label track titled Transcription appears below the audio waveform. Each label contains a phrase from the audio, synchronized to the waveform by timestamp.

Step 4: Export the Transcript from Audacity

1. Click anywhere in the label track to select it
2. Aller à File > Export Other > Export Labels
3. Choose your format:
   • SRT: for video editors and YouTube captions
   • WebVTT: for web-based media players
   • Plain text (.txt): for documents, notes, and search
4. Name the file and save it

The exported file contains every phrase with its start and end timestamp, making it ready to drop into a video editor or upload as a caption file.

Method 1 at a Glance

• Coût : Gratuit
• Plugin download: Several GB (base model: under 1 GB; full install with all models: 6 GB+)
• Processing: On-device (your hardware)
• Langues : 99
• Identification des locuteurs : Experimental (English-only, 2-speaker max)
• Output: Label track exported as SRT, VTT, or plain text
• Privacy: Fully local: audio never leaves your computer
• Idéal pour : Single-speaker clean audio, offline, air-gapped environments

Method 2: Cloud Transcription With Sonix (99% Accurate)

This is the most reliable method when you need to transcribe Audacity recordings automatically at scale. It delivers higher accuracy across multi-speaker recordings, supports 53+ langues, and produces transcripts ready for legal documentation, content repurposing, or team sharing, with full Diarisation des locuteurs par IA and export to 30+ formats. Sonix is the leading cloud transcription platform trusted by Google, Stanford, and ESPN.

Step 1: Export Your Audacity Recording

If you want to work with a processed, edited version of your recording rather than uploading the raw AU project file, export from Audacity first:

1. In Audacity, go to File > Export Audio
2. Choose your format: MP3 (compressed, smaller file) or WAV (uncompressed, largest quality)
3. Set the quality settings and name the file
4. Cliquez sur Exportation and save to your computer

Alternatively, you can upload Audacity’s native .AU format directly to Sonix without any export step: Sonix accepts AU files alongside all common audio and video formats.

Step 2: Upload Your Audio to Sonix

1. Go to sonix.ai and sign in (or start your 30-minute free trial: no credit card required)
2. Cliquez sur Télécharger on the Sonix dashboard
3. Select your Audacity audio file from your computer, or import directly from Google Drive or Dropbox if you store recordings there
4. Sonix accepts MP3, WAV, M4A, AU, FLAC, OGG, AAC, and most other audio and video formats

Step 3: Select Language and Start Transcribing

1. After uploading, Sonix prompts you to select the langue spoken in the recording
2. Choose from Plus de 53 langues prises en charge: Sonix automatically handles technical vocabulary, proper nouns, and domain-specific terminology across all of them
3. Cliquez sur Start Transcribing

Sonix processes files up to 10x faster. A 60-minute Audacity recording can return a complete transcript in under five minutes, depending on conditions.

Step 4: Review, Edit, and Export

Sonix opens the finished transcript in its synchronized editing interface. Every word is linked to the corresponding audio timestamp: click any word to jump directly to that moment in the recording.

Key features available in the Sonix editor:

• AI speaker diarization: Sonix automatically detects speaker changes and labels each speaker’s dialogue throughout the transcript. For a two-person interview, this means every exchange is cleanly separated without manual tagging
• Word-level timestamps: every single word carries a millisecond-accurate timestamp, useful for subtitle syncing and legal documentation
• In-browser editing: correct any word in the synchronized editor and the change applies across all export formats simultaneously
• Search: use Ctrl+F to find any word or phrase instantly within a transcript of any length

To export your finished transcript:

1. Cliquez sur le bouton Exportation button in the top-right of the editor
2. Choose your format:
   • DOCX or PDF: for reports, documentation, and sharing
   • SRT or VTT: for video captions (YouTube, Premiere Pro, Final Cut, DaVinci Resolve)
   • Plain text: for note-taking apps, CMS uploads, or AI tools
   • 30+ additional formats: including JSON with timestamps, for developers and data pipelines

Sonix's integrations page lists all direct connections, including Adobe Premiere Pro, Final Cut Pro, and Hindenburg Journalist, which are common workflows for Audacity users who move audio into professional editing tools.

Method 2 at a Glance

• Users: 6.2M+ users (Sonix-reported)
• Coût : $10/hr standard rate; subscription plans from $25/mo (additional hours billed at $10/hr on all plans)
• Essai gratuit : 30 minutes, aucune carte bancaire requise
• Précision : jusqu'à une précision de 99%
• Langues : 53+
• Identification des locuteurs : Full AI speaker diarization across all plans
• Output: 30+ formats: DOCX, PDF, SRT, VTT, JSON, and more
• La sécurité : SOC 2 Type II certified; AES-256 encryption; HIPAA-ready via Medical Sonix (BAA available)
• Idéal pour : Multi-speaker recordings, professional output, enterprise compliance

Which Method Fits Your Workflow?

Both methods let you transcribe Audacity recordings automatically. The right one depends on your accuracy needs, volume, privacy requirements, and whether you need features like speaker diarization or multi-format export. In our evaluation of both options, Sonix consistently outperforms the local plugin on accuracy and output quality, while the OpenVINO Whisper plugin is the best free, fully offline choice.

Audacity OpenVINO Whisper Plugin:

• Coût : Gratuit
• Précision : Good for clean audio
• Langues : 99 (Whisper base model)
• Identification des locuteurs : Experimental (tdrz model)
• Vitesse de traitement : Depends on your hardware
• Export formats: SRT, VTT, TXT
• Editing interface: Audacity label track
• Privacy / offline: Fully local, no uploads
• Idéal pour : Single-speaker, offline, no budget

Sonix Automated Transcription:

• Coût : $10/hr standard rate; subscription plans from $25/mo
• Précision : jusqu'à une précision de 99%
• Langues : 53+
• Identification des locuteurs : Full AI speaker diarization
• Vitesse de traitement : Up to 10x faster than real time (under five minutes per hour of audio)
• Export formats: 30+ formats: DOCX, PDF, SRT, VTT, TXT, JSON, and more
• Editing interface: Synchronized browser editor
• Privacy / offline: Cloud-based; SOC 2 Type II certified; AES-256 encryption; HIPAA-ready via Medical Sonix
• Idéal pour : Multi-speaker, high accuracy, professional output

For solo podcasters, researchers with one-time recordings, or anyone working in a fully air-gapped environment, the built-in OpenVINO method handles the task without any external service. For teams, content workflows, journalism, legal documentation, or any project where speaker labeling and export flexibility matter, Sonix delivers the accuracy and workflow features that raw Audacity label tracks do not offer.

Common Mistakes to Avoid

• Not denoising before transcribing. Background noise, even a subtle hum or room reverb, significantly affects transcription accuracy in both methods. Always apply noise reduction in Audacity before exporting or running Whisper. Audacity’s built-in Noise Reduction effect (Effect > Noise Reduction) and the OpenVINO AI Denoising effect both work well; the AI denoising is more effective for recordings with heavy interference.
• Uploading a multi-track Audacity project without mixing down. Audacity projects (.AUP3) can contain multiple tracks, including different microphone channels, music beds, and sound effects. Before transcribing, make sure you’re working with a single mixed-down audio file. Use Tracks > Mix > Mix and Render in Audacity before exporting.
• Choosing the wrong language setting. In both methods, manually specifying the recording language produces more accurate results than relying on auto-detection, especially for short clips, technical vocabulary, or recordings that start with silence. Always select the language explicitly.
• Skipping the speaker diarization step for multi-person recordings. If your Audacity recording contains an interview, panel discussion, or meeting with multiple speakers, using a method without AI speaker diarization produces a single undivided block of text that is hard to read and difficult to attribute. Enable diarization in the Whisper settings or use Sonix, which applies speaker diarization automatically on every transcript.
• Exporting in the wrong format for the downstream use case. SRT files are for video captions: don’t paste them into a Word document. Plain text strips timestamps: don’t use it for subtitle imports. Match the export format to the destination: SRT or VTT for video, DOCX or PDF for documentation, TXT for plain-text workflows.

Tips for Better Automatic Transcription Results

Pre-process audio before transcription.

The noise reduction techniques in Audacity, including noise gates, equalization, and compression, directly improve automated transcription accuracy. A 30-second noise profile capture followed by noise reduction and a gentle high-pass filter (to cut low-frequency rumble) is the standard pre-processing workflow for interview recordings. Apply these effects before transcribing with either method.

Use Initial Prompt for brand names and technical terms.

 Whisper (via the Initial Prompt field) and Sonix handle proper nouns and technical vocabulary more accurately when given a short context hint. For a product interview, entering the company name, product names, and industry terms as an initial prompt prevents common misspellings in the final transcript.

Set a custom vocabulary for recurring terminology.

If you transcribe Audacity recordings regularly in a specialized domain such as medical, legal, or engineering, Sonix’s custom vocabulary feature lets you define terms that should always be transcribed a specific way, preventing inconsistencies across a large batch of files.

Batch upload for high-volume workflows.

If you record daily Audacity sessions, such as podcast interviews, client calls, or field audio, Sonix supports batch uploads. Drop multiple files at once, and all of them are processed in parallel. Sonix’s API also supports automated upload pipelines for teams that need transcripts delivered to a database or CMS without manual steps.

Leverage timestamps for content repurposing.

Every Sonix transcript includes word-level timestamps. For podcast editors, this means you can search the transcript for a specific quote, note the timestamp, and jump directly to that moment in Audacity to make precise cuts without scrubbing through the entire recording.

Note on translation.

Sonix's fonctionnalité de traduction converts a completed transcript into 54+ langues directly in the platform. Translation is available and billed separately at the same hourly rate as transcription.

Verdict final

There is no single best method for every Audacity workflow. Here is how to decide when choosing how to transcribe Audacity recordings automatically:

• Pour offline transcription of clean, single-speaker recordings, Audacity’s OpenVINO Whisper plugin is the right fit. It is free, fully private, and requires no external service. It handles solo podcast sessions, lecture captures, and any workflow where audio never leaves your machine. The OpenVINO plugin is the only fully local transcription option built into Audacity.
• Pour multi-speaker recordings, professional output, or high-volume workflows, Sonix is the most complete option. It delivers 99% accurate, AI speaker-labeled, fully formatted transcripts in under five minutes with no label track cleanup, no manual speaker tagging, and export-ready for Word, PDF, YouTube, or your video editor. Manual transcription typically requires 4 to 6 hours per hour of audio; Sonix automates this in minutes, as detailed in Sonix’s transcription resources.
• Pour teams in healthcare, legal, or media that need enterprise-grade security alongside automated transcription, Sonix is the top choice. Sonix is SOC 2 Type II, HIPAA-ready via Medical Sonix (BAA available), and encrypts all files with AES-256 encryption. Organizations including Stanford, Google, and ESPN use it for sensitive recordings.

If your primary need is fast, accurate, formatted transcripts from Audacity recordings without the setup overhead of a large plugin install, Sonix is the most complete path.

Next Steps

Both methods covered in this guide automate the manual effort of transcribing audio. For quick offline transcription of single-speaker clean audio, the OpenVINO plugin handles the job without any external service. For high-accuracy transcription of multi-speaker recordings, Prise en charge de plus de 53 langues, AI speaker diarization, and export to the formats your workflow actually needs, Sonix is the faster and more complete path.

Now that you know how to transcribe Audacity recordings automatically, choose the method that fits your setup and start saving hours of manual transcription work in 2026.

Essayez Sonix gratuitement: 30 minutes, aucune carte bancaire requise.

Explore Sonix’s full feature set, including translation into 54+ languages, automated summaries, and direct integrations with Premiere Pro, Final Cut Pro, and Audacity.

Questions fréquemment posées

Can Audacity transcribe audio to text natively?

Yes. With the free OpenVINO AI Plugin installed, Audacity can transcribe audio to text directly inside the app using OpenAI’s Whisper model. The transcript appears as a time-stamped label track. This feature is available for Windows, macOS, and Linux (via the OpenVINO AI plugin). Use the Audacity version recommended by the OpenVINO plugin release notes; current releases are in the 3.7.x line.

What audio formats does Sonix accept from Audacity?

Sonix accepts Audacity’s native .AU format directly, along with MP3, WAV, M4A, AAC, FLAC, OGG, and most other common audio and video formats. You can upload directly from your computer, or import from Google Drive or Dropbox.

How accurate is Sonix automatic transcription?

Sonix livre jusqu'à une précision de 99% à travers 53+ langues. Audacity’s built-in Whisper plugin also performs well on clean, single-speaker recordings. Accuracy in both methods decreases with background noise, overlapping speakers, or strong accents, which is why denoising before transcription is strongly recommended regardless of which method you use.

Does Sonix support speaker diarization?

Yes. Sonix automatically applies AI speaker diarization to every uploaded recording. It detects speaker changes and labels each speaker’s dialogue throughout the transcript with no manual setup required. Audacity’s Whisper plugin also includes experimental speaker diarization via the small.en-tdrz model, though it supports English only, with a maximum of two speakers.

Is Sonix secure enough for sensitive recordings?

Yes. Sonix is SOC 2 Type II, HIPAA-ready via Medical Sonix (BAA available), and encrypts all files with AES-256 encryption at rest and in transit. Organizations in healthcare, legal, media, and research, including Stanford, Google, and Adobe, use Sonix to transcribe sensitive audio.