Telegraf Studio: a breaking story is found, drafted, and published. Illustration built on the real interface.
01
Editorial problem
I built this system from scratch, from product design to deployment. It now runs in two national newsrooms in Ukraine. The screens here come from Telegraf Studio, the version made for Telegraf.com.ua.
A news article is more than its text. Before it goes out, someone has to add context, pick a search headline, write the description, translate the piece, and fill in the publishing form. Each of those steps used to live in a different tool.
The editor puts them in one window, in the order an editor does them.
02
Five steps
01Text
02Add-ons
03SEO
04Translation
05Export
03
Add-ons
The second step is where an article gets richer. Background research finds sources and prepares context. The video tool turns the piece into slides and scenes. The collage builder makes roundup covers.
Three image tools cover the rest. One generates covers and illustrations, one edits an image that already exists, and one translates the text inside an infographic.
Add-ons next to the draft: background, video, collages, and three image tools.
04
SEO
The SEO step starts from a real search query. The editor picks a keyword from Serpstat data and the system writes several headline variants around it.
Every variant shows the keyword it is built on and a length meter with the target range, so the editor can see why one is recommended. The same step produces the on-page headline, the description, and tags.
Search headline variants built on a Serpstat keyword, each with a length meter.
05
Translation
Telegraf publishes in Ukrainian and Russian. The translation step puts both versions side by side with the same structure, images, and captions, and the translated side stays editable.
Original and translation side by side. The article text is left in the original languages.
06
Export
Before export the editor sees a short checklist: both language versions, the main photo, the search headline. Anything missing links back to the step that fixes it.
The article then goes to the newsroom's own CMS through its API, as a draft, for immediate publication, or on a schedule.
Where to aim the next article is a separate product. DiscoRadar shows what Google Discover is recommending across the country, and it has its own case study.
07
Editorial throughput
One workflow
Writing, media, SEO, translation, and export stay in the same workspace, in a fixed order.
Visible operations
Provider use, failures, failover, and estimated AI cost are logged, so the newsroom can see what its AI production costs.