Halaman ini masih berbahasa Inggris.
Process
Drop a PDF, CSV, or image and get structured output back.
Jalaran Process is a file-processing toolbox that turns an uploaded PDF, spreadsheet, image or transcript into something you can actually use: a summary, a caption, structured notes or a converted document.
Who Process is for
For the moment you have a file and a small, specific job to do with it (summarise this report, merge these four PDFs, turn this transcript into notes) and do not want to hand the document to a free web converter of uncertain provenance to get it done.
What Process does
Process is the module that actually contains tools, in the ordinary sense of the word: a grid of them, each doing one thing to one file. Merge PDFs, extract text, summarise a document, caption an image, convert Markdown to PDF, generate a DOCX, summarise a CSV, structure meeting notes from a transcript, or ask questions of a document you uploaded. Jobs run in a thread pool off the request path, uploads are validated by their actual magic bytes rather than by a trusted filename, and files live in a private bucket that no browser session can read directly.
- PDF merge, text extraction, summarisation and Markdown-to-PDF
- Image captioning, CSV summarisation and DOCX generation
- Meeting notes structured from a pasted or uploaded transcript
- Magic-byte upload validation into a private, service-key-only bucket
How Process works
Pick a tool
The grid lists what is available right now, including which AI-backed tools the model server can currently serve.
Upload the file
Validated by magic bytes, not by extension, and stored in a private bucket rather than anywhere a link could reach.
Let the job run
Work happens in a thread pool off the API event loop, so a slow PDF parse never blocks anything else you are doing.
Download, or ask follow-up questions
Take the output file, or for document Q&A keep asking questions against the same uploaded document.
What Process does not do
Audio transcription is reported unavailable rather than quietly failing: the model server in this build exposes no working transcription endpoint, so the tool tells you so instead of returning an empty result. Meeting notes therefore work from a transcript you supply, not from a recording. Process is quota-limited per day (3 jobs a day on the free plan), and it transforms files rather than training on them; anything that learns from your data is Analyze. Outputs are purged on a retention schedule.
Common questions
- Why does audio transcription say it is unavailable?
- Because it genuinely is. Jalaran’s model server in this build accepts text and images but rejects audio, and its transcription endpoint does not respond, so the tool reports Unavailable rather than returning a plausible empty transcript. Meeting Notes still works from a pasted transcript.
- Is my uploaded file readable by anyone else?
- No. Files go into a private bucket that is write-only to the backend service key, with a storage policy denying direct access, and they are delivered back to you through short-lived signed URLs.
- How many files can I process per day?
- There is a per-day job quota: 3 jobs a day on the free plan. AI-backed tools also draw on the shared daily AI quota, so the limit is enforced in the backend rather than by hiding the module.
Works well with
Drag a dataset in for real ML: classification, regression, clustering and anomaly detection.
Read research PDFs, annotate passages, and ask questions across your documents.
Transcribe a recording on Jalaran’s own servers, then extract the decisions and the action items.
Ask a component datasheet a question and get the exact page it came from.
Text-to-speech via Kokoro on Jalaran’s own servers, carrying a phoneme timeline inside the file.
Animate a character image so it speaks your audio, rendered on Jalaran’s own GPU.