Go back

What costs tokens during data ingestion?

Understand what types of data ingestion cost tokens and why

What types of content consume tokens during data ingestion?

Processed Data ingestions consume tokens as they require additional processing (such as parsing, structuring, extraction, diarization, transcription, or frame analysis) beyond plain-text syncing

How are documents priced in tokens?

Documents cost 5 tokens per page (or equivalent). The “equivalent” depends on the document type, but the logic is consistent: you're paying for the amount of content Astell has to process and index.

• Supported formats: PDF, .docx, .pptx, .xlsx, Google Docs/Sheets/Slides, Pages/Numbers/Keynote
• How pricing is calculated:
  • PDFs: 5 tokens per page
  • Word / Google Docs: 5 tokens per page
  • PowerPoint / Slides: 5 tokens per slide
  • Excel / Sheets: 5 tokens per sheet (tab)

Examples:

• 10-page PDF: 50 tokens
• 25-slide deck: 125 tokens
• 3-sheet workbook: 15 tokens
• 100-page manual: 500 tokens

How are images priced in tokens?

Images cost 1 token per image, regardless of file size.

• Supported formats: JPEG/JPG, PNG, GIF, SVG, WebP, HEIC, BMP, TIFF
• Examples: screenshot in Slack (1), product photo in email (1), diagram in Notion (1)
• Profile pictures: 0 tokens because they aren't processed for search

How do embedded images in documents affect token cost?

If a document contains embedded images, Astell processes those images separately in addition to the document itself. You pay the document rate plus 1 token per embedded image.
Example: a 10-slide presentation with 5 embedded photos costs (10 × 5) + (5 × 1) = 55 tokens.

How are videos priced and what's included in video processing?

Videos cost 8 tokens per minute. Video processing includes automatic transcription and frame analysis so the AI can understand what's being said and what's happening visually.

• Supported formats: MP4, MOV, AVI, WebM, MKV
• Examples: 5-min demo (40), 30-min recording (240), 2-hour video (960)
  Tip: for long recordings, consider extracting audio only or processing just the key segments.

How are audio files priced and what's included in audio processing?

Audio costs 2 tokens per minute and includes automatic transcription. Once transcribed, the text becomes searchable.

• Supported formats: MP3, WAV, M4A, OGG, FLAC
• Examples: 10-min snippet (20), 60-min meeting audio (120), 5-min voice memo (10)