Skip to content

Talkr v0.1.6

Choose a tag to compare

View all tags
@github-actions github-actions released this 05 Oct 11:56
v0.1.6
23d1a8a

New

  • Transcription quality in Settings: Fast, Accurate, or Auto (the default), which is
    accurate for imported files and fast for recordings made in Talkr. Accurate makes fewer
    mistakes and takes about 1.5 times as long.
  • Save speech as MP3, FLAC or WAV. Talkr remembers the format you used last.
  • Save transcripts as plain text, Markdown, SRT or WebVTT subtitles, CSV or JSON.
  • Update notifications: Talkr checks for new versions at launch and every few hours, shows what
    is new, and lets you install now, later, or skip a version. Settings has a "Check now" button
    and a switch to turn automatic checks off.
  • Playback speed (0.75× to 2×) in every player.
  • In History, click a timestamp to play from that point, copy a transcript with timestamps, or
    open speech in Speak to hear it with another voice.
  • Speak shows the word count and roughly how long the speech will be, and keeps what you typed
    when you switch screens.
  • Press ? for a list of keyboard shortcuts. Shortcut hints show ⌘ on macOS.

Fixed

  • Long transcriptions on slower computers are no longer stopped as "not responding".
  • Cancelling a queued job no longer cancels the job that is running.
  • Running out of memory on the GPU no longer turns the GPU off for good.
  • Switching to the CPU stops the GPU engine right away, freeing video memory.
  • Leaving Transcribe mid-recording now keeps and transcribes the recording.
  • A double click on Play no longer marks the audio as missing.
  • After switching speech models, Generate waits for the new model's voices.
  • The update banner no longer covers the page's own buttons.
  • A failed model list now offers Retry instead of saying nothing is installed.
  • Error messages stay while you point at them, so they can be read and copied.
  • macOS now asks for microphone access the proper way.
  • Model names that end in a dot or match a Windows device name are refused.

Free, private, offline speech tools. Text to speech with Kokoro and Piper, speech to text with Whisper, all running on your own computer.

Speech to text uses your graphics card when it can (Vulkan on Windows and Linux, Metal on macOS) and falls back to the CPU on its own. The engines run in a separate process, so a model that runs out of memory can't take the app down.

Which file do I need?

  • Windows: *_x64-setup.exe (recommended) or *_x64_en-US.msi
  • macOS Apple Silicon: *_aarch64.dmg
  • macOS Intel: *_x64.dmg
  • Linux: *.AppImage, *.deb or *.rpm

Builds are not code-signed yet, so Windows SmartScreen and macOS Gatekeeper will warn on first launch.
Checksums are listed in SHA256SUMS-*.txt.

Assets 23
Loading