All features

A transcription engine runs on your computer. Real-time transcription during microphone recording is available on every plan. Pro also provides the live view while recording computer audio or a mix of microphone and computer audio. You can also transcribe files you already have.

Quality

Choose High quality (large-v3), Balanced (large-v3-turbo), or Fast (medium). You can also select a transcription model directly.

Language

Detected automatically, or you can specify one of the major languages yourself.

Real time

Every plan can show text while recording with a microphone. Pro also supports the live view for computer audio and mixed microphone and computer audio. When recording stops, post-recording transcription starts.

Formats it reads

Thirteen formats: mp3, wav, m4a, flac, ogg, aac, opus, wma, mp4, mov, mkv, webm, and avi.

On Pro, after you stop recording or import a file, the transcript can be separated into as many as 4 speakers. It also works with audio captured on a single microphone.

Once separated, speakers can be reassigned in the editor.

On Pro, the AI reads the transcript and shapes it into what you need. All AI inference, including summaries, document creation, and chat, runs on your computer, without cloud AI.

Working with text

Summaries, minutes, decisions, action items, bullet lists, keywords, titles, Q&A, proofreading, translation, emails, blog drafts, and tone analysis.

Working with transcripts and conversations

Search and read the current transcript and conversation, and inspect transcript statistics.

Long-term memory

Save reusable knowledge on your computer and retrieve it in later conversations.

Producing documents

Create spreadsheets, charts, images, CSV, Word, Excel, PowerPoint, and PDF files.

Past output

Search past output, copy it into the current workspace, and reuse it.

Editing suggestions

Transcript changes arrive as suggestions. By default, you review and approve each change before it is applied. You can also choose to apply all changes automatically.

Separate AI tasks

Start a separate AI task, send follow-up instructions, and check its progress.

Record from a microphone. Pro can also record your computer's audio, the microphone and computer together, or audio from a specific app. No recording bot has to join the meeting.

Audio sources

A microphone; on Pro, computer audio, both together, or per-process capture to record audio from a specific app.

Requirements

Per-process capture is available on Windows 11 / Windows Server 2022 or later, Linux with a running PipeWire session, and macOS 14.4 or later. Recording on Linux requires PipeWire and ALSA.

Importing from YouTube (Pro)

Import audio or video from an HTTPS YouTube URL. tranext connects to YouTube when you start the import.

The transcript is yours to correct. Fix the boundaries, reassign a speaker, and listen back as you go.

Segments

Split and merge as needed.

Speakers

Reassign who said what.

Undo

Undo and redo are available for transcripts recorded or imported in the app.

Playback

Change the playback speed and read the waveform.

Export in a shape the next task can use — as subtitles, as prose, or as a table.

Text formats

Nine formats: SRT, VTT, TXT, JSON, LRC, CSV, TSV, ASS, and Markdown.

Audio and video

Audio can be exported in six formats: WAV, FLAC, MP3, AAC, OGG, and Opus. For video, copy the original without changing its format or extract its audio.

Adding a subtitle track to video (Pro)

Add subtitles to a video as a subtitle track while leaving the picture unchanged. Exporting subtitle files such as SRT is available on every plan.

Supported platforms: Windows (x64), Linux (x64), macOS (Apple Silicon). Transcription and AI on Pro use an available GPU automatically, falling back to the CPU when needed. Pro speaker separation uses the GPU through WebGPU and falls back to the CPU when the GPU is unavailable.

Windows

Windows 11 or later (x64) is supported. Windows 10 may run the app, but some features are not guaranteed.

Linux minimum requirements

Launching the app and transcribing files on the CPU on Linux x64 require glibc 2.35 or later, such as Ubuntu 22.04, Debian 12, or Fedora 36 and later. RHEL 9 (glibc 2.34) is not supported.

All features on Linux

Using all features requires the minimum requirements plus GPU drivers (NVIDIA 560.35.03 or later, or a Vulkan driver). Recording requires PipeWire and ALSA; running the AppImage requires FUSE 2.

GPU

Transcription and AI on Pro automatically select CUDA, Vulkan, or Metal, falling back to the CPU when needed. Pro speaker separation uses the GPU through WebGPU and falls back to the CPU when the GPU is unavailable.

Network

The model catalog is fetched from engines.sadola.dev at startup and every 24 hours while the app runs. There is no dedicated setting to stop this connection. App updates are checked once at startup; you can disable this in settings. Engine updates are checked and installed automatically at startup; you can disable the check and automatic installation separately. When needed, models are downloaded from Hugging Face, and llama and yt-dlp from their official GitHub releases. When you search for an AI model, the search query is sent to Hugging Face. License activation happens when you request it, with offline activation also available. When you start a trial, the app connects to api.sadola.dev. Notifications are fetched when you open them, and the license is re-checked only after that fetch. Feedback is sent only when you press Send; its body is what you write. Importing from YouTube (Pro) connects only when you request it. There are no automatic crash reports, usage statistics, or diagnostics uploads.