Log Transcription
Photograph handwritten growing logs and extract structured data automatically using AI
adminLog Transcription
The Log Transcription page lets you photograph a handwritten growing log and use AI vision to extract structured data — dates, weights, strain names, observations, contamination notes, and more — without manual data entry. It lives at Logs > Transcribe.
Role requirements
You must have at least the Manager role to use log transcription. Viewers and Workers will see an access-denied message.
How it works
The wizard walks you through four phases. A step indicator at the top shows your progress.
Phase 1 — Capture
Upload or drop an image of your handwritten log:
- Click the drop zone or drag a file onto it to select an image from your device.
- Supported formats: JPG, PNG, HEIC, WebP. Photos taken directly on a phone work well.
- After selecting an image, a preview is displayed. Click the X button on the preview to choose a different image.
Lifecycle Stage selector: Choose the stage the log was recorded at (for example, Fruiting or Harvesting). Providing the stage helps the AI focus on fields relevant to that stage and improves extraction accuracy.
Click Transcribe once you have selected an image and a stage.
Phase 2 — Processing
Sporeline uploads your image to secure storage and sends it to the AI vision model for analysis. A spinner is shown while this is in progress. This step typically takes a few seconds.
Phase 3 — Review
The review screen shows your original image alongside the extracted data.
Extraction Summary shows:
- The number of log entries found in the image
- How many you have confirmed so far
- The AI’s overall confidence level for the extraction (High, Medium, or Low)
Warnings (if any) are shown as amber banners. These flag fields the AI was uncertain about — pay particular attention to those fields when reviewing.
Entry cards: Each entry identified in the log is shown as a card with editable fields:
| Field | Description |
|---|---|
| Strain | Mushroom strain name |
| Stage | Lifecycle stage noted in the log |
| Log Date | Date the log was written |
| Inoculation Date | Date of inoculation if noted |
| Logged By | Staff member name if recorded |
| Harvest Weight | Weight and unit, for example 250 g |
| Spawn Lot | Spawn lot identifier |
| Spawn Supplier | Spawn supplier name |
| Observations | Free-text observations and notes |
You can edit any field before confirming. Fields the AI extracted are pre-populated but are not locked — correct any mistakes directly in the inputs.
For each entry, either:
- Click Confirm to accept the data (the card turns green), or
- Click Skip to discard that entry (the card dims and fields are hidden).
Use Confirm All to accept every entry at once.
Click Finish to complete the review. The button shows the number of confirmed entries and is disabled until every entry has been either confirmed or skipped.
Phase 4 — Done
The results screen shows how many entries were confirmed. If all entries were skipped, nothing is saved.
Click Transcribe Another to start a new transcription from the beginning.
Tips for best results
- Good lighting matters more than camera quality. Take photos in a well-lit area. Avoid harsh shadows across the log.
- Flat, unobstructed pages work best. If the log is in a spiral notebook, flatten the page before photographing.
- One page at a time. Submit one log page per transcription for the most accurate extraction. Multi-page spreads may produce lower confidence results.
- Select the correct stage. The stage selector is the most impactful accuracy lever — a mismatched stage may cause the AI to look for the wrong fields.
- Always review before confirming. The AI is accurate but not infallible. Dates, weights, and lot numbers especially warrant a quick visual check against the original image shown on the left.
AI model note
Log transcription is powered by an AI vision model. Extraction quality depends on handwriting legibility, image quality, and whether the page content matches the selected stage. Confirmed entries are not automatically applied to block records — you will need to apply them manually to the relevant blocks after reviewing.