About this release
This is the transcription component for Premiere Pro, the piece that listens to the dialogue on a sequence and writes out a time aligned transcript you can turn into captions. It matters as a separate download because the language models are what actually do the recognition, and having the full offline set installed is the difference between transcription that runs on the machine with the network off and one that will not start at all.
The workflow inside Premiere is short. Point it at a sequence or a clip, pick the spoken language, and it produces a transcript panel where every word is linked to its moment on the timeline. From there the transcript becomes a caption track that you restyle in the Essential Graphics panel, position with safe margins for each platform, and burn in or export as a sidecar file. Speaker labelling separates a two hander so the captions read as a conversation.
Because the recognition runs locally against these models, longer footage simply takes proportionally longer rather than depending on a connection or a service quota. Keeping the component matched to the Premiere build is the thing to watch, since a transcription feature that greys out is almost always a missing or mismatched language pack, which is exactly what this package supplies.
What it does
- Local speech recognition with the full offline language model set
- Time aligned transcript linked word by word to the timeline
- One step conversion of a transcript into an editable caption track
- Speaker labelling to separate dialogue between people
- Caption styling through the Essential Graphics panel
- Safe margin presets for captions per delivery platform
- Sidecar caption export in the common subtitle formats
- Search across the transcript to jump to a spoken phrase
- Support for the languages listed for Premiere transcription
- Runs with the network off once the models are installed
Changes in this version
- Added and refreshed several transcription language models
- Improved accuracy on overlapping dialogue and crosstalk
- Faster transcription pass on longer sequences
- Better speaker separation on two person conversations
- Fixed a case where the feature greyed out after a Premiere update
What the package contains
- Adobe Speech to Text component and full language model set
- Offline models for all supported transcription languages
- Readme with the install order written out step by step
- Notes on matching the component to the Premiere build
- Sample sequence showing a finished caption pass
Installation notes
Follow these in order. Most reports of a failed install come from skipping a step rather than from the package itself.
- Close Premiere Pro before you start.
- Disconnect from the network.
- Extract the archive to a short path such as C:\Setup.
- Run the component installer and let the language models unpack.
- Open Premiere, run a short transcription to confirm it works, then close it.
- Reconnect to the network.
System requirements
| Operating system | Windows 10 version 22H2 or Windows 11 |
| Host application | A matching Premiere Pro build must be installed |
| Processor | Intel or AMD 64-bit multi core processor |
| Memory | 16 GB recommended for long sequences |
| Graphics | GPU with 4 GB VRAM helps the pass run faster |
| Storage | 2 GB free for the language models |
Things worth knowing
- This is a component for Premiere, not a standalone application, it has no window of its own.
- The models have to match the Premiere version, a mismatch greys the transcription feature out.
- Longer footage takes proportionally longer, the pass runs on the machine rather than a service.
Questions about this title
Is this a standalone transcriber?
Does it send my audio anywhere?
Why is transcription greyed out in Premiere?
Which languages are covered?
What people are saying
Transcription came back after this fixed the language pack. Two hundred clips captioned in an afternoon.
Runs with the network off, which is the whole reason I use the local version. Solid.
Speaker separation on an interview was cleaner than I expected. Saved a lot of manual splitting.
Listing information
This entry is catalogued under Video & Motion and carries the tags below. Language coverage is Multilingual, covers the supported transcription languages, and the supported platform list is Windows 10, Windows 11. Version 2.2.5 is the current catalogued build, last checked 4 days ago, and the entry has been in the index since it was added 1 year ago.