feature release

· Ryan Joseph Kamp

reword-nerd v0.8: Image prompt packages

At a glance

v0.8 adds an orange Image companion beside the established Text workbench. Images remain in the browser while the new workflow prepares one reference-image prompt and provider run card per included source, with focused improvements to document review and both package formats.

Added

Changed

Fixed

Why this matters

Reference-image prompting benefits from the same inspectable preparation that already supports text rewriting. Source custody, model-specific instructions, requested changes, warnings, and export hashes stay visible, while repeated settings can be applied across a selected group without losing the ability to refine one image at a time.

See it in action

The Image portal includes its own silent Quick Start for the complete working flow. The separate silent, synthetic v0.8 walkthrough embedded with this post highlights the companion portal and the broader polish without substituting for that tutorial.

How to use it

Choose Image in the header, add supported images or a folder, focus and review each source, and select any group that should share explicit settings. Review warnings and optional OCR, confirm the image set, then build the package in memory. Inspect the package preview before deliberately downloading the timestamped ZIP. Use Text for document rewriting, including Continuous or Gallery PDF review when a source has multiple pages.

Compatibility and limitations

The Image portal accepts PNG, JPEG, WebP, and AVIF directly and recovers supported visuals from bounded PDF, DOCX, folder, and ZIP inputs. Unsupported encodings, nested archives, links, encrypted containers, unsafe paths, and limit violations are rejected with local messages. Image-generation output remains stochastic; faces, visible text, logos, fine geometry, and structured layouts require human review. Midjourney guidance is best-effort variation, and provider capabilities can change after a profile's recorded verification date.

Privacy

Images, extracted visuals, OCR, filenames, settings, prompts, and packages remain in the active browser session. No model is called, no credential is requested, no provider charge is incurred, and no selected source is uploaded by reword-nerd. Exact source-byte preservation may retain EXIF or location metadata inside exported images, so packages should be reviewed before sharing. Public tutorial and Updates media use synthetic same-origin assets and remain separate from session content.

Verification

Verification covers Image intake signatures and limits, archive safety, reducer and confirmation revision behavior, prompt profiles, deterministic package bytes and hashes, clipboard and drag fallbacks, object-URL disposal, Text fixture parity, static Updates rendering, responsive containment, accessibility, privacy, and direct Pages-base routes. Chromium coverage is recorded without claiming unrun cross-browser behavior.

What comes next

Post-release verification remains distinct from future marketing media and external writing. Those materials remain independent later work rather than part of the product package.

Feedback and contribution links

Synthetic Updates walkthrough poster for reword-nerd v0.8: Image prompt packages
Silent, synthetic release walkthrough. Read the transcript.

Video unavailable? Open the release poster.

← All Updates