> ## Documentation Index
> Fetch the complete documentation index at: https://docs.goframe.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Transcribe media

> Turn speech in video and audio into a word-level transcript with speakers: settings, languages, statuses, cost and what a transcript unlocks

A transcript is the text of what is said in a clip, with a timestamp for every word and a label for every speaker. Transcribe a file once and the same transcript powers text-based editing, captions and the agent's understanding of your footage.

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/transcription-settings.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=9d729e849cc50671d3fb1efa8194f5f9" alt="Transcription settings popover with Model, Language, Speaker detection and Keep filler words" width="832" height="1150" data-path="images/editor/transcription-settings.png" data-path="images/editor/transcription-settings.png" />
</Frame>

## Start a transcription

Any of these starts the same job:

* Open the **Transcript** panel in the **Editor** layout. Each video or audio file in the project is listed with a **Transcribe** button; before the first transcript the panel reads **Edit video by editing text — Transcribe this file to cut, trim and clean it up straight from the transcript.**
* In the **Captions** panel, click **Transcribe N files** to transcribe every clip with speech on the Timeline at once.
* Ask the agent. Prompts such as "Add captions" or "Remove filler words" transcribe the footage first if needed.

<Note>
  A transcript belongs to the asset, not to a clip. Cut, split or duplicate the clip as much as you like; every piece keeps its words.
</Note>

## Transcription settings

Before the job starts you can adjust **Transcription settings** (the sliders icon in the Transcript panel header). As the popover says, they apply to new transcriptions; use **Re-transcribe** on a file to apply new settings to it.

| Setting               | Options                                                | Notes                                                                                                                                          |
| --------------------- | ------------------------------------------------------ | ---------------------------------------------------------------------------------------------------------------------------------------------- |
| **Model**             | **Highest accuracy (Best quality)** / **Multilingual** | Highest accuracy is the default and covers the 18 languages below. Multilingual supports 99 languages; pick it for anything outside that list. |
| **Language**          | **Auto-detect** or a specific language                 | Choose the language explicitly when auto-detection picks the wrong one, for example in a bilingual recording.                                  |
| **Speaker detection** | on / off                                               | "Label who is speaking." On by default, so you can rename and reassign speakers later.                                                         |
| **Keep filler words** | on / off                                               | "Transcribe "um"/"uh" — needed for one-click filler removal." On by default.                                                                   |

### Languages

**Highest accuracy** offers: English, Español, Français, Deutsch, Italiano, Português, Nederlands, Dansk, Suomi, Norsk, Svenska, Türkçe, العربية, עברית, हिन्दी, 日本語, 中文, Tiếng Việt.

**Multilingual** adds: Русский, Українська, Polski, Čeština, Română, Ελληνικά, Magyar, 한국어, Bahasa Indonesia, ไทย. Its Auto-detect mode recognizes 99 languages in total.

<Tip>
  Leave **Keep filler words** on for talking-head footage. Filler removal in the Transcript panel only works on words that were transcribed; if you switch it off, the fillers are silently dropped from the text but remain in the audio.
</Tip>

## Statuses

Transcription runs in the background. The clip on the Timeline shows a **Transcribing** badge, and the file's row in the Transcript panel and the **Background tasks** panel show the current stage:

| Status                       | What is happening                                                                                                                             |
| ---------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------- |
| **Queued for transcription** | The job is registered.                                                                                                                        |
| **Extracting audio**         | The audio track is separated from the video in your browser, with a percentage and elapsed time (for example **Extracting audio 5% · 0:01**). |
| **Uploading audio**          | The audio is sent for transcription.                                                                                                          |
| **Transcribing**             | Speech recognition is running.                                                                                                                |
| **Transcript ready**         | Done. The Transcript panel fills in and captions turn on if the Captions panel was waiting (**Transcript ready — captions turned on**).       |
| **No speech detected**       | Audio was found but no words were recognized.                                                                                                 |
| **Transcription failed**     | Something went wrong. Use **Re-transcribe** to try again.                                                                                     |

If the file is not on the Timeline yet, the panel then reads **Add this file to the timeline to view and edit its transcript.** Click **Cancel transcription** in the Transcript panel to stop a running job.

<Frame>
  <img src="https://mintcdn.com/goframe-fb816310/nCYzUacADHcqMX4R/images/editor/transcription-status.png?fit=max&auto=format&n=nCYzUacADHcqMX4R&q=85&s=7c7c90696852fa5c1b044c66e8e5faee" alt="Timeline clip with a Transcribing badge and the Background tasks panel showing the stage" width="832" height="1150" data-path="images/editor/transcription-status.png" data-path="images/editor/transcription-status.png" />
</Frame>

## Cost and limits

* Transcription costs about 1 credit per minute of audio. Credits are charged once per asset; editing the transcript and building captions from it cost nothing extra.
* A source file can be up to 24 hours long.
* Up to three transcriptions run at the same time; beyond that, extra jobs wait for a free slot.

See [Plans and credits](/editor/plans-and-credits) for how credits work.

## Re-transcribe

Use **Re-transcribe** on the file in the Transcript panel to run the job again, for example after changing the language or model, or when a partial transcript needs fixing. Re-transcribing replaces the transcript and is charged again.

<Warning>
  Re-transcribing replaces the previous transcript. Check speaker names, text corrections and caption word-timing adjustments afterwards, as they were made on the old one.
</Warning>

## What a transcript unlocks

* **Text-based editing**: delete a sentence in the Transcript panel and the clip is cut on the Timeline. See [Edit video by editing text](/editor/transcript-editing).
* **Captions**: the Captions panel builds every caption page from the transcript's words and timings. See [Captions](/editor/captions).
* **Search**: find any spoken phrase with **Search transcript**.
* **Agent context**: the agent reads transcripts to find highlights, remove repeated takes, cut by topic, or reference a passage you `@`-mention in chat.

## Troubleshooting

* **No speech detected**: check the audio in Preview. Music-only or very quiet tracks will not transcribe. Try again with the language set explicitly.
* **Transcription failed**: re-transcribe from the panel. If it keeps failing, check the file's status in the **Background tasks** panel and that you have credits left.

## Continue

* [Edit video by editing text](/editor/transcript-editing)
* [Captions](/editor/captions)
* [Organize the Media pool](/editor/media-pool)
* [Troubleshooting](/editor/troubleshooting)