Transcribe Audio — Editor’s Look Docs
Make the caption file from the sequence’s own audio, on your machine, and check every line before it is placed.
- **Auto-detect and Sinhala.** The model listens for the language itself. Sinhala is the one language it can mistake, so if you know the speech is Sinhala, choose **සිංහල** yourself. If at least a tenth of the speech turns out to be Sinhala, an Auto-detect run switches to the Sinhala model by itself, English words stay English, and the captions are treated as Sinhala, font and all (**Checking the language…** shows on the cover, with **Cancel**). It does not do this with **Translate to English** on. If the Sinhala model is not downloaded, a red card says *Sinhala heard in the audio*: download it (choose සිංහල in Transcribe and press its **Download**), or press **Use the current model anyway** to carry on with the model you chose, which will write Sinhala poorly. **OK** or Esc stops the run with nothing changed. A run that switched models does not change your saved choice.
#### Sinhala
Whisper’s own models do not write Sinhala well, so Sinhala has a model of its own, **Sinhala** (488 MB): SinhaSpeech’s Sinhala fine-tune of Whisper small, converted for whisper.cpp, the engine Transcribe runs. **English words stay in English letters**, so speech that mixes the two comes out as mixed captions.
- **Choose Sinhala as the language and only this model is offered**; any other language hides it. **Create captions** waits until it is downloaded.
- **It comes from our own copy only**, at `dl.editors-look.com`, and is checked like every model.
- **The audio is transcribed in short pieces**, of at most 8 seconds, cut at its quietest moments and put back in place. The percentage covers them all, and **Cancel** stops the one running.
- **Every word keeps its own time.** Where the model gives a run of words the same time at the end of a piece, those words are spread out, so no two captions start together and none is lost. The words around them keep their aligned times.
**Use a Unicode Sinhala font for the captions.** An older Sinhala font (FMBasuru and the like) draws English letters as Sinhala. While the Text box holds Sinhala, MogrtCaptions links to a set of Unicode Sinhala fonts — see **Sinhala text and fonts** in **Styling controls & presets**.
Launch offer: every plugin is 25% cheaper than its list price until the end of 30 November 2026 (UTC).
- DropSnap — Screenshots, photos and graphics onto your timeline in one click — transparency baked in, background removed, without leaving Premiere. Free, with every feature. No account needed.
- MogrtCaptions — Transcribe your sequence’s own audio into motion-graphic captions across the whole timeline. In the mogrt template you chose. Free edition, never expires; Pro is $24 list price (Launch offer: $18 until 30 November), one-time payment — Buy Once, Lifetime License. 10 free Pro runs.
- Stockpile — Search five free stock libraries from one panel. Organised by what you need rather than by which site has it. Preview it, then put it straight on the timeline. $52 list price (Launch offer: $39 until 30 November), one-time payment — Buy Once, Lifetime License. 14-day free trial.
- Edge — Dialogue-driven edits in one Premiere panel: take out the dead air, and cut a multicam podcast from who is talking. $64 list price (Launch offer: $48 until 30 November), one-time payment — Buy Once, Lifetime License. 14-day free trial.
This site needs JavaScript to show each page in full. A plain-text summary is at /llms.txt.