Overview
Send a signed work order photographed on a phone, a stack of supplier invoices off an office scanner, or a handwritten intake form into an automation. The Convert File to Text step reads the words off the page and hands them forward as plain text. A page that is an image rather than typed characters is read the same way as any other page, so a photo of paper behaves like a document.
TL;DR: The Convert File to Text step reads scanned PDFs, PNG, JPG, WebP, and HEIC files and returns plain text. Point it at a photographed work order or an office-scanner invoice, and the next step in your automation gets the words. Build an automation →
What you need to do
Nothing. There is no toggle, no setting, and no separate step to add.
Point the Convert File to Text step at your file the same way you point it at a typed PDF. When the page is a scan or a photo, the words are read out of the image and returned as text. That reading is optical character recognition (OCR), and it runs on its own as part of the step.
┌──────────────────────────────────────────────────────────────┐
│ Automation › Step 2 › Convert File to Text │
├──────────────────────────────────────────────────────────────┤
│ File URL [ @Form Submission · Attachment ] │
│ │
│ Reads typed pages · scanned pages · photos │
│ │
│ Output Plain text → available to every later step │
└──────────────────────────────────────────────────────────────┘
What the Step Reads
| What you send | What the step does |
|---|---|
| Scanned or image-only PDF | Reads the words off the page |
| PNG, JPG / JPEG, WebP | Reads the words off the page |
| HEIC / HEIF, the format an iPhone camera makes | Reads the words off the page |
| GIF, BMP | Not covered by this path |
| Word (DOCX), PowerPoint (PPTX, PPT), Excel (XLSX, XLS), EPUB | Read directly as text. Nothing special needed |
| CSV, TXT, Markdown | Read directly as text. Nothing special needed |
Keep a scan or a photo under 20 MB. That is the ceiling for reading words off an image. The formats in the last two rows are read as text rather than off an image, so this ceiling does not apply to them and your plan's upload size is the limit instead.
That covers the two shapes paper arrives in. A page of typed characters is read straight through. A page that is a picture of characters, which is what a scanner and a phone camera both produce, is read off the image.
What it uses
Reading the words off an image uses AI credits, the same as any other AI step in an automation. A document that already has selectable text is read straight through and does not use them, so a typed PDF costs you nothing extra. See AI Credits and Models for how credits work across your workspace.
Where this works
Inside the Convert File to Text step in automations. That is the surface, and it is worth being precise about it, because two nearby paths behave differently.
- The Convert Media to Project step brings media into a project. It is a separate step with its own behavior, and reading a scan is not part of it.
- Asking an AI agent a question about a file you uploaded runs through agent knowledge and the Media Tab. That is a different path from this step.
So when you want the words off a scan, put a Convert File to Text step in the automation and read from there.
One thing worth knowing if you have published an app. A file someone picks in a form field in your Taskade Genesis app is sent to the automation as a real stored file, and it lands in your workspace's Media Tab where you can see it. The same holds for files sent to your app's webhook. See File Uploads in Your App for where those files live and what reaches them.
Four Ways Operators Use It
Field service. A tech photographs a signed work order on site. Point the Convert File to Text step at that photo and the next step gets the customer name, the address, and the sign-off date as text it can write into a job row. See Invoice Automation for the billing half of that loop.
Bookkeeping. A month of supplier invoices comes off the office scanner as image-only PDFs. Run each one through the step, then hand the text to Ask Agent and ask for the vendor, the invoice number, and the total. Finance apps shows what to do with the rows afterwards.
Recruiting. Candidates send resumes as phone photos and scanned printouts. The step returns the text, and a Categorize with AI step sorts them by role before anyone opens a file. Applicant screening walks the rest of the pipeline.
Clinic intake. A front desk scans a completed intake form. The step reads what is on the page and passes it forward, so the record is built from the paper rather than retyped from it. Pair it with form logic when the follow-up questions depend on the answers.
Common Patterns
| You want to | Do this |
|---|---|
| Read a photographed work order | Point the Convert File to Text step at the photo |
| Pull totals out of scanned invoices | Convert to text, then Ask Agent for the fields you want |
| Sort scanned resumes by role | Convert to text, then Categorize with AI |
| Teach an agent from a scanned manual | Convert to text, then Add Knowledge to Agent |
| Start the run when paperwork arrives | Use a form, webhook, or scheduled trigger |
Related guides
- Convert File to Text. The step that does the reading, with its full settings
- Ask Agent. Hand the text to an agent and get named fields back
- Automation Triggers. Start the automation the moment a file arrives
- File Uploads in Your App. Where files from your published app land
- Document Intake App. A full build with a document as the input
- Agent Knowledge. Teach an agent from what the step read
- Build automations. Browse automations you can drop a reading step into