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

# Transcription

> Generate accurate transcripts from video and audio with Wikio's AI-powered speech-to-text, then search, edit, and export them.

Wikio uses AI-powered speech-to-text to turn spoken audio into searchable, editable transcripts. Transcripts make your content discoverable, enable subtitling, and provide written records of every conversation.

## Generate a transcript

<Steps>
  <Step title="Open an asset">
    Navigate to the video or audio file you want to transcribe from your library or project.
  </Step>

  <Step title="Run the transcription workflow">
    Click **Workflows** and select **Transcribe**. Wikio detects the spoken language automatically, or you can choose it manually.
  </Step>

  <Step title="Wait for processing">
    Transcription runs in the background. You'll receive a notification when it's ready. Processing time depends on the length of your file.
  </Step>

  <Step title="Review the transcript">
    The transcript appears in the **Transcript** panel alongside the player. Click any segment to jump to that point in the media.
  </Step>
</Steps>

<Callout type="info">
  You can transcribe files in over 50 languages. For best results, make sure the audio is clear and the correct language is selected.
</Callout>

## Review and edit

Click any text segment to edit it directly. Changes save automatically.

* **Fix errors**: Correct names, technical terms, or misheard words.
* **Adjust timing**: Drag segment boundaries in the timeline to realign text with speech.
* **Split or merge segments**: Use the toolbar icons to divide a long segment or combine short ones.

Edits apply to the transcript only and do not alter the original media file.

## Speaker identification

Wikio can detect and label different speakers in a conversation.

1. After transcription, open the **Speakers** section in the transcript panel.
2. Assign names to each detected speaker (e.g., "Speaker 1" becomes "Sarah").
3. Speaker labels appear throughout the transcript and carry over to exports.

<Callout type="info">
  Speaker detection works best with two to six speakers. Overlapping speech may reduce accuracy.
</Callout>

## Multi-language transcription

For content with multiple spoken languages, you can run transcription once per language or let Wikio auto-detect language switches within the same file.

* Each language generates a separate transcript track.
* Switch between language tracks using the dropdown in the transcript panel.

## Export transcripts

Export your transcript for use outside Wikio.

1. Open the **Transcript** panel and click **Export**.
2. Choose a format:

| Format   | Description                                              |
| -------- | -------------------------------------------------------- |
| **TXT**  | Plain text without timecodes—ideal for articles or notes |
| **SRT**  | Timed segments for use as subtitles in video players     |
| **DOCX** | Formatted document with speaker labels and timestamps    |
| **JSON** | Structured data for integrations and custom workflows    |

3. Click **Download**.

## Use transcripts for search

Transcribed content is automatically indexed. You can search across all transcripts in your workspace using the global search bar.

* Search returns matching segments with timestamps so you can jump directly to the right moment.
* Filter search results by project, team, or date to narrow down results.

See [Search history](/features/search) for more on finding content.

## Best practices

* **Transcribe early**: Run transcription as soon as assets are uploaded so content is searchable from the start.
* **Review proper nouns**: AI handles common speech well, but always check names, brands, and jargon.
* **Use speaker labels**: Labeling speakers makes transcripts easier to read and more useful for meeting notes.
* **Choose the right export format**: Use TXT for quick sharing, SRT for subtitles, DOCX for formal records.

## Troubleshooting

* **No transcript generated**: Ensure the file has audible speech. Music-only or silent files produce empty results.
* **Low accuracy**: Check that the correct language is selected. Background noise, heavy accents, or overlapping speakers can affect quality.
* **Missing speakers**: If some speakers aren't detected, check that their audio is loud enough relative to other speakers.