Lana 0.1.2 (preview)

A preview of Lana: Praat's acoustic analysis with word, phone and speaker tiers filled by speech recognition, exported as a TextGrid. Please report anything that goes wrong under Issues.

What's new since 0.1.0

New

• Updates from inside the app. From this version on, Lana tells you when a new version is out and installs it with one click. On a Mac, "Open Anyway" is needed only for this first install.
• Language by name. One Language box for the recording, searching some 1,200 languages by name, code or everyday name: "Taiwanese" or "Hokkien" finds Min Nan, "Farsi" finds Persian. Each model is sent the code it understands, and a model that does not know the language says so.
• A queue for runs. Run or ⟲ while a model is working adds the job to a queue, with progress and a cancel button for each.
• Tiers fill as the model works, a 30-second stretch at a time, for every built-in model.
• Figures and measurements for papers. ⬇ Figure / measurements… exports the panes at 300 or 600 dpi (PNG) or as SVG, with the Praat settings as a caption, and saves tables of f0, formants, duration and intensity (CSV).
• Right-click menus on intervals, tiers, the waveform, the spectrogram, the segment table and the transcript.
• Models folder. Settings shows where downloaded models are stored, how much space they take, and can move them to another folder or disk.
• Easier zooming. The overview strip shows which part of the recording is on screen, and each tier counts the intervals outside the window.
• Cloudflare Workers AI as an online recogniser, with a free daily allowance of about three and a half hours of Whisper large-v3 turbo.

Fixes

• MMS for Mandarin, Cantonese, Japanese and Korean ("cannot pad lm_head.bias to 512"). Chinese and Japanese get one interval per character.
• Recordings that play but would not transcribe ("unsupported codec"): any file that plays can now be transcribed.
• Whisper on Windows and Linux no longer crashes on processors without AVX-512, and runs about seven times faster on Windows.

Which file to download

System	File
macOS, Apple Silicon (M1 or later)	Lana_0.1.2_aarch64.dmg
Windows 11 (10 untested)	Lana_0.1.2_x64-setup.exe
Linux, Debian or Ubuntu	Lana_0.1.2_amd64.deb
Linux, other distributions (including Arch)	Lana_0.1.2_amd64.AppImage

Intel Macs are not supported in this version.

Opening it for the first time

Lana is not yet signed with an Apple or Microsoft certificate, so both systems warn before its first launch, once.

macOS. Open the .dmg and drag Lana into Applications. Open Lana; macOS will say it cannot verify the app. Click Done, then open System Settings → Privacy & Security, scroll down to the message about Lana, and click Open Anyway.

Windows. Run the installer. If "Windows protected your PC" appears, click More info, then Run anyway.

Linux. On Debian or Ubuntu, sudo apt install ./Lana_0.1.2_amd64.deb (updates from inside the app need the AppImage). Elsewhere, make the .AppImage executable and run it; it needs fuse3.

MMS is licensed for non-commercial use only; Lana asks you to accept that licence before downloading it.

Lana 0.1.0 (preview)

The first public preview of Lana: Praat's acoustic analysis with word, phone and speaker tiers filled by speech recognition, exported as a TextGrid. This is an early version. Please report anything that goes wrong under Issues.

Which file to download

System	File
macOS, Apple Silicon (M1 or later)	Lana_0.1.0_aarch64.dmg
Windows 11 (10 untested)	Lana_0.1.0_x64-setup.exe
Linux, Debian or Ubuntu	Lana_0.1.0_amd64.deb
Linux, other distributions (including Arch)	Lana_0.1.0_amd64.AppImage

Intel Macs are not supported in this version.

Opening it for the first time

Lana is not yet signed with an Apple or Microsoft certificate, so both systems warn before its first launch. The warning concerns who published the app, not anything the app does, and it appears only once.

macOS. Open the .dmg and drag Lana into Applications. Open Lana; macOS will say it cannot verify the app. Click Done, then open System Settings → Privacy & Security, scroll down to the message about Lana, and click Open Anyway. Confirm with your password. From then on Lana opens normally.

Windows. Run the installer. If a blue window says "Windows protected your PC", click More info, then Run anyway.

Linux. On Debian or Ubuntu, sudo apt install ./Lana_0.1.0_amd64.deb. Elsewhere, including Arch, make the .AppImage executable (chmod +x) and run it; it needs fusermount3, from the fuse3 package, which most desktops already have.

What it does

• Waveform, spectrogram, pitch, formants and intensity computed by Praat itself, with every analysis setting on screen.
• Word, phone (IPA) and speaker tiers, filled by recognition models that run on your own computer, offline once downloaded. Choose and download them under ⚙ Connection in the toolbar.
• Hand annotation, TextGrid import and export, and recording from a microphone. Annotation is saved as you work.
• Online recognition services can be added, with keys kept in your system's keychain: OpenAI, Groq, and Cloudflare Workers AI, whose free daily allowance covers about three and a half hours of audio with Whisper large-v3 turbo.

The built-in models include Whisper, ZIPA (IPA phones in any language), Omnilingual ASR and MMS. MMS covers more than 1,150 languages but is licensed for non-commercial use only; Lana asks you to accept that licence before downloading it. See the README for which model suits which language.