> For the complete documentation index, see [llms.txt](https://evidano.gitbook.io/ailyze-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://evidano.gitbook.io/ailyze-docs/prepare-research-data/convert-images-and-pdfs-to-text.md). # Convert Images & PDFs to Text Turn photos, scans, and PDFs of pages into editable, searchable text, test the prompt on one page first, then correct and download the results. This page shows you how to use the **Image to Text** workflow in full: which files it takes, how languages and domains shape the result, how to test before you run a whole batch, and what you can do with the text afterwards. It is for anyone whose data exists as pictures of pages: handwritten field notes, completed paper questionnaires, letters, archive scans, whiteboard photos, or screenshots. For a short first run, start with [Convert Your First Image to Text](/ailyze-docs/getting-started/convert-your-first-image-to-text.md); for a list of every control, see [Image to Text Settings](/ailyze-docs/reference/settings/image-to-text.md). ## Before you start * An Evidano account. The first pages are free, then each page is charged from your wallet; see [Plans, Limits & Credits](/ailyze-docs/reference/plans-limits-and-credits.md). * Page images or PDFs. Accepted formats and caps are listed in [Supported File Types](/ailyze-docs/reference/supported-file-types.md). * Pages you can read yourself. Transcription cannot recover text the camera did not capture. ## Step 1: Add images or PDFs From your account page, click **Image to Text**. Every upload starts a new project; to finish setting up pages you uploaded earlier, open that project and click **Continue setup**. One project holds either images or PDFs, never both. If you drop a PDF into an image project, Evidano asks you to remove the current files first. * **Images** go into a single project. Upload from the **Local Files** drop zone or the **Google Drive**, **Dropbox**, and **OneDrive** tabs. * **PDFs** each become their own project, named after the PDF, and Evidano splits each one into a page image per page. Every page counts as an image for your allowance and for billing. A row reads "Preparing pages…" until the split finishes; if **Next Step** reports that the server is still splitting, wait a moment and click it again. Files upload the moment you add them, and the project exists from that point. You can change the **Project name** until you leave this step. Hover a thumbnail to **Preview file** or **Remove file**, then click **Next Step**. ## Step 2: Choose a language and a domain The **Language** column lists the languages Evidano has prompts for. The **Domain** column lists the kinds of material within that language. Together they set the instructions the model follows, so the right pair matters more than anything else you do here. * **Historical hands.** Languages such as English, Latin, German, French, Spanish, Arabic, and Ottoman Turkish list dated palaeography domains, for example **Tudor and Jacobean Secretary, Italic, and Mixed Hands, c. 1500–1650**. Rows with an info icon have reading guides: click it to open the palaeography guides, a list of external references for that hand. * **Modern material.** English adds everyday domains: **Handwritten Content**, **Forms and Questionnaires**, **Scanned Documents**, **Business and Financial Documents**, **Screenshots and Digital Captures**, **Educational Materials**, **Whiteboards and Workshop Materials**, **Signs, Labels, and Printed Materials**, **Official and Identity Documents**, and **Tables and Structured Content**. Other languages offer **General**. * **Auto-detect.** At the top of the language list, for mixed or unknown languages. No domain is needed; the text comes back in the original language. * **Custom.** Choose **Custom** as the language to type one that is not listed, or **Custom** as the domain and describe the pages under **What should the model expect on these pages?**. Your description is added to the prompt. The search box at the top matches language names, domain names, and the wording of each domain's instructions, so searching for "invoices" or "Kurrent" finds the right row.
Reading guides dialog listing reference links for an early modern English secretary hand

Each guide is an external reference for reading that hand. Open them before you transcribe difficult manuscripts.

Once the domain is set, the box headed **ADDED TO YOUR PROMPT** shows the exact instructions the domain contributes. At the bottom, **Sample page** names the page the next step will test. Click **Test a different page** to choose another page, or **Pick at random**. Open any page full screen to read it first. The sample starts transcribing in the background as soon as the language and domain are set.
Sample page picker showing uploaded page thumbnails with a Pick at random button

Pick the hardest page as the sample. If the prompt copes with it, it copes with the rest.

## Step 3: Test the prompt on one page Click **Next Step**. The review step shows the sample page beside its transcription. Sample runs are free, do not use your allowance, and leave the other pages untouched. The **Transcription prompt** is plain text you can edit: add instructions such as "expand 'Resp.' to 'Respondent'" or "keep the two-column layout", then click **Transcribe again**. **Change page** tests a different page. Read the result against the image, then choose: * **Transcription looks good, proceed with this prompt for all N files**: every page runs with this prompt and the same model as the sample you approved. * **Transcription looks bad**: click **Edit prompt and try again** (edit, then **Transcribe again**, or **Keep this result** to go back to the cards), or **Retry with a more advanced AI model**. The advanced model handles faint, cursive, or damaged pages better at a higher price per page. Approving an advanced sample runs the advanced model on every page. * **Skip this sample transcription and transcribe all images**: run every page with the prompt as written and the basic model, using the **Start transcription** button. {% hint style="warning" %} **Costs money:** When a run goes beyond your free allowance, the **Confirm image transcription charge** dialog shows how many pages are covered, how many will be charged, the rate for the model you chose, and your balance before and after. Large batches also offer a smaller paid test batch first. Nothing is charged until you confirm. See [Plans, Limits & Credits](/ailyze-docs/reference/plans-limits-and-credits.md). {% endhint %} After you confirm, Evidano queues every page and opens your project list. Transcription runs in the background, and the list updates itself while pages are transcribing. There is no email for this workflow. ## Step 4: Review and correct each page The **Image to Text** list shows every page as a card. The **Projects** sidebar lists your projects; click one to open it. Switch between **Grid view** and **List view**, and search by **Project & file name** or inside the **Transcribed text** itself.
Image to Text list with the projects sidebar, search bar, and a grid of page cards

The search scope menu switches between file names and the transcribed text, so you can find a page by a word written on it.

Open a page to see it beside its transcription. The viewer offers **Fit to view**, zoom levels, **Rotate**, **Download** for the original image, and three layouts: **Split view**, **Image only**, and **Text only**. **Formatted** shows tables and headings as on the page; **Raw** shows the source text with line numbers. **Files** opens a panel listing every page in the project.
Viewer with the manuscript image and zoom controls beside the text pane with Formatted, Raw, Edit, Copy, DOCX and XLSX

Use Previous page and Next page, or the arrow keys, to move through the project.

* **Edit** opens the text for correction, and **Save** keeps your changes. * **Mark verified** records that you checked the page. Verified pages are labelled in the list, so you can see what still needs a read-through. * **Retry** on a **Failed** page's card in the list runs it again with the advanced model. * **Delete file** removes one page, and **Delete all files** in the files panel removes every page. ## Step 5: Export the text Per page, click **Copy**, **DOCX**, or **XLSX** in the text pane. **DOCX** keeps headings, lists, and tables, with the page image; for a page from a PDF, choose **This page** or **Whole PDF**. **XLSX** is built from a table on the page, or holds the text when there is none. Per project, **Spreadsheet** in the project header downloads a workbook with one row per page. In bulk, tick pages in the list to reveal a selection bar with **Download image**, a **Download N transcripts** menu (separate DOCX files, one DOCX, or one spreadsheet), **Delete**, and **Clear selection**. Several pages download together as a ZIP. The exported files are ordinary documents. To find themes, click **Use in a document project**; upload the spreadsheet to a spreadsheet project when each page is one respondent. ## Next steps * Turn the exported pages into themes and quotes: [Analyze Interview Transcripts](/ailyze-docs/common-research-workflows/analyze-interview-transcripts.md). * Pages in another language? Transcribe them in that language, then [Translate Files](/ailyze-docs/prepare-research-data/translate-files.md). * Unreadable output, a stuck page, or a charge you did not expect: [Image to Text Issues](/ailyze-docs/troubleshooting/image-to-text-issues.md).