Google quietly widened what developers can feed into the Gemini API and what tools can send back. The November 5, 2026 entry in the Gemini API release notes lists two related changes: broader Microsoft Office and OpenDocument file support on the input side, and multimodal function responses that let tools return Office files, audio and video instead of only text, images and PDFs.
On the input side, the API now accepts PowerPoint 97-2003 (PPT), OpenDocument Presentation (ODP) and OpenDocument Spreadsheet (ODS) files, and it adds page-image processing for Word 97-2003 (DOC) files. DOC, PPT and ODP files are rendered as page or slide images so the model can read layouts, charts and pictures. For ODS workbooks the model receives the data and formulas from the first sheet only. Processing for the modern formats that were already supported (DOCX, DOTX, ODT, PPTX, XLSX and XLS) is unchanged.
Google's document-processing guide spells out how each format is handled. DOCX, DOTX, ODT and PPTX files are converted to full-document text, with embedded images kept for DOCX and PPTX, but pages and slides are not rendered as images, so layout and charts can be lost. XLSX and XLS still expose only the first sheet as CSV-like cell values without formulas or formatting. ODS is the exception among spreadsheets: the first sheet arrives with both cell data and formulas. Other sheets in a workbook are ignored unless the developer moves them to the first position or splits them into separate files. Existing file-size and page limits still apply to the whole file.
The second change is aimed at tool-using agents. Function responses can now carry Office and OpenDocument files, audio and video back to the model. In the GenerateContent API, function response parts accept those media types in addition to images, PDFs and plain text; they also accept fileData so a tool can return a Files API URI instead of inline bytes. displayName is available on both inlineData and fileData, and video parts take a mediaProcessing setting of STATIC or AGENTIC. In the Interactions API, function_result steps accept document, audio and video content blocks (alongside text and image), by inline data or URI. Video blocks are available in the v1beta and v1alpha channels and accept a processing field.
The multimodal function-calling documentation ties the feature to Gemini 3 series models. Supported content block types include text, image (PNG, JPEG, WebP), document (PDF, CSV and the Office and OpenDocument types listed above), audio in the formats already documented for Gemini, and video in the formats already documented for video understanding. Each media block can use base64 data or a Files API URI, and files returned by tools are processed the same way as the same file passed as ordinary input. An unsupported MIME type returns a 400 INVALID_ARGUMENT error that names the field and lists the accepted types.
Taken together, the two notes close a practical gap for document-heavy agents. A tool can now fetch a legacy PPT deck or an ODS sheet, hand it straight back to Gemini without a separate conversion step, and keep going in the same conversation. The release is still a documentation-only changelog: Google published no benchmark, pricing change or customer example with the November 5 entry, and no independent coverage appeared during this run.
Developers who already send DOCX and PPTX files will see little change on that path. Teams that still receive DOC, PPT or OpenDocument files from partners, regulators or older enterprise systems get a first-party way to keep those formats inside the Gemini loop, including when the files arrive as tool results rather than as the user's original upload.