Overview
This is the language model component that gives the timeline editor its transcript panel. On a normal install the models arrive through the application manager, and when they do not, the transcript panel sits there empty and nothing explains why. Installing the component directly fixes that and puts transcription entirely on the machine.
Once installed, any clip or sequence can be transcribed in place. The result is a script pinned to timecode, split by speaker, with filler words detected and marked. Searching the transcript jumps the playhead, and deleting a sentence in the text performs the matching ripple edit on the timeline with handles preserved. For interview, documentary and podcast work that removes the slowest part of an assembly.
The captions side runs from the same transcript. Generate a caption track, choose line length and duration rules, restyle every caption at once from the graphics panel, and burn them in or export a sidecar file in the standard subtitle formats. Because the models are local, none of the audio leaves the machine, which is the usual reason a facility keeps this component installed rather than working through a service.
What it does well
Local transcription
Language models run on the machine, so media never leaves the building and transcription works with the network disconnected.
Speaker labelling
Speakers are separated automatically and can be renamed once across the whole project rather than clip by clip.
Text based editing
Delete a line in the transcript and the sequence ripples to match, with pad handles kept for retrimming later.
Caption generation
Build a caption track from the transcript with configurable line length and duration, restyled globally from the graphics panel.
Subtitle export
Write sidecar files in the common subtitle formats, or burn captions into the picture at export.
Changes in this build
- Recognition improved for accented speech and for overlapping speakers on a shared microphone.
- Filler word detection extended to more languages in the bundled set.
- Faster transcription on multicore machines through better batching.
- Caption line breaking reworked to avoid splitting short phrases awkwardly.
- Fixed a failure when transcribing sequences containing offline media.
What is in the package
- Speech to text component installer, 64-bit
- Offline language models for the supported languages
- Caption formatting presets
- Subtitle export templates
- Installation notes for placing the models manually
System requirements
| Processor | Intel or AMD 64-bit, six cores or more recommended |
| Memory | 16 GB recommended when transcribing long sequences |
| Graphics | Any GPU supported by the host editor |
| Storage | 6 GB free for the model set |
| Display | 1920 x 1080 minimum |
| System | Windows 10 or Windows 11, 64-bit, with a matching editor generation installed |
Installation
- Close the timeline editor completely before starting.
- Extract the archive to a local folder.
- Run the installer, which places the models where the editor expects them.
- Reopen the editor and check that the transcript panel lists the installed languages.
- Transcribe a short clip once to confirm the models load correctly.
Before you start
The component has to match the editor generation, a mismatched pair leaves the panel empty.
Model files are large, do not install them onto a small system partition.
If the language list is blank, the model folder was moved after install, run setup again.
Questions about this title
Does transcription need a connection?
No, the models are local and run entirely on the machine.
Which languages are covered?
Eighteen language models ship in this package and appear in the transcript panel after install.
Does it caption automatically?
Captions are generated from the transcript in one step, with styling and timing rules you control.
Can I export subtitles separately?
Yes, sidecar files in the common subtitle formats, or burned in at export.
About this listing
This entry was checked on a clean install of Windows 10 / 11 (64-bit) before it was published, and it is rechecked whenever the package is rebuilt. The figures on this page come from the site index rather than from the publisher, so the download count is what people here have actually pulled.
If a mirror stops answering, use the contact form and it gets replaced. Reports about a specific title are handled faster than general messages because they name the file, the mirror and the point at which the transfer stopped.
Requests for a different version, a different language build or an older release go on the requests page. Older versions of a title are usually still held even when only the current one is listed.