V-GEN Builds Full HD TV News Clips with Gemini and FFmpeg, No Video Editor Required

Habr has featured V-GEN, a local service with a web interface that produces news clips end to end. It searches for news hooks in live mode, generates the story with Google Gemini, assembles the video with FFmpeg and outputs a finished Full HD file with a Russian filename for a broadcast playlist. The announcement gives no figures on speed, cost or clip length; the details are hidden under the cut. • The whole cycle, from topic search to finished file, runs without manual editing • Gemini handles the text, FFmpeg assembles the video • The main question is fact-checking before broadcast

AI-processed from Habr AI; edited by Hamidun News
V-GEN Builds Full HD TV News Clips with Gemini and FFmpeg, No Video Editor Required
Source: Habr AI. Collage: Hamidun News.
Listen to article
Article Fact Sheet & Key Figures
ParameterDetails
Primary sourceHabr AI ↗
Editor-in-Chief
Key metric #1The whole cycle, from topic search to finished file, runs without manual editing
Key metric #2Gemini handles the text, FFmpeg assembles the video
Key metric #3The main question is fact-checking before broadcast

V-GEN is a local service with a web interface that takes a news clip through the entire cycle on its own: it searches for news hooks in live mode, prepares the story and outputs a Full HD file with a Russian filename that can go straight into a broadcast playlist. It runs on Google Gemini and FFmpeg. The project is described on Habr, but the announcement itself is short: it contains no data on speed, cost or clip length. For now, that means judging the architecture, not benchmarks.

How the pipeline works

The logic can be read from the description. Gemini takes the tasks that require working with language: picking topics from a current feed, writing the text and structuring the story. FFmpeg handles the mechanics: stitching, overlays and encoding at 1920×1080. The web interface serves as a control panel through which the operator starts and monitors the process.

The service is local. It runs on the user's machine and reaches out only for data and to the model API. For a small newsroom, this is convenient: no subscription to a cloud video editor is needed, and the pipeline stays under the team's control.

One separate detail is the Russian name of the output file. It is a small thing, but it reveals the target audience: people who have broadcast automation waiting for files under a clear naming scheme. This is a tool for production, not a demo for social media.

What is known so far

  • Format: local service with a web interface
  • Model: Google Gemini
  • Video: FFmpeg
  • Output resolution: Full HD
  • Input: live search for news hooks
  • Output: a finished file with a Russian filename for a broadcast playlist
  • Not disclosed: generation time for a single story, query cost, clip duration, image source, voiceover setup

The absence of these items matters. They determine whether V-GEN is fit for broadcast or remains an experiment.

Where the pitfalls are

The main risk is accuracy. A language model entrusted with both topic search and text can distort a figure, mix up a date or attribute a quote to the wrong person. In a regular newsroom, the duty editor and a fact-checker do this work. If V-GEN has no manual review step, errors will go to air along with the finished file.

The second question is the visuals. A news story without footage turns into a slideshow, and using other people's footage raises rights issues. The announcement does not say where the images come from.

The third is broadcast requirements. Television is sensitive to codec, frame rate and loudness level. Full HD does not mean the file will pass a particular broadcaster's acceptance check.

Finally, "local" does not mean "independent of the cloud." Requests to Gemini go to Google, and that is worth keeping in mind if stories are based on confidential materials.

Who it may suit and how to test it

Potential users include small regional channels, corporate TV, screens in lobbies and public transport, and newsrooms with regular digests. Automation pays off most wherever volume matters more than distinctive authorial style.

The practical testing procedure is simple:

  • Run 10–15 stories and check every fact against the primary source
  • Measure time and token cost per clip
  • Check the file parameters against the requirements of your broadcast system
  • Keep a human as the duty editor: V-GEN prepares a draft, and the editor decides what goes on air

If the author reveals pipeline details in the full article, the project has a chance to become a handy template for building such a pipeline yourself with open tools.

Story

Key Questions Answered

How the pipeline works?

The logic can be read from the description. Gemini takes the tasks that require working with language: picking topics from a current feed, writing the text and structuring the story. FFmpeg handles the mechanics: stitching, overlays and encoding at 1920×1080. The web interface serves as a control panel through which the operator starts and monitors the process. The service is local. It runs on the user's machine and reaches out only for data and to the model API. For a small newsroom, this is convenient: no subscription to a cloud video editor is needed, and the pipeline st

What is known so far?

Format: local service with a web interface Model: Google Gemini Video: FFmpeg Output resolution: Full HD Input: live search for news hooks Output: a finished file with a Russian filename for a broadcast playlist Not disclosed: generation time for a single story, query cost, clip duration, image source, voiceover setup The absence of these items matters. They determine whether V-GEN is fit for broadcast or remains an experiment.

Where the pitfalls are?

The main risk is accuracy. A language model entrusted with both topic search and text can distort a figure, mix up a date or attribute a quote to the wrong person. In a regular newsroom, the duty editor and a fact-checker do this work. If V-GEN has no manual review step, errors will go to air along with the finished file. The second question is the visuals. A news story without footage turns into a slideshow, and using other people's footage raises rights issues. The announcement does not say where the images come from. The third is broadcast requirements. Television is sen

ZK
Zhemal Khamidun — CPO at Alpina Digital, product manager of AlpinaGPT. Building AI products since 2019. Hamidun News is a pet project that grew into a production aggregator with 6,000+ readers.
✦ Unified AI Workspace

Access Frontier LLMs & Multimodal AI in One Workspace

Discover NomadNet AI (nomadnet.ai) — unified access to frontier language models, research synthesis, and multimodal workflows.

Try NomadNet AI →Launch NomadNet Agent
What do you think?
Loading comments…