Operational deep-dive · Published 2026-05-06 · By Vertical Visuals
How a Prague Agency Ships 2 TV Episodes Per Week
A 6-person Prague agency delivers 73 broadcast TV episodes per year to a national sports broadcaster. This is the operational deep-dive on how that pipeline runs — batching, specialization, AI tooling, and the structural choices that make weekly broadcast cadence sustainable without a 50-person team.
The contract by the numbers
- Volume: 73 broadcast TV episodes per year (29 weekly highlights + 44 studio show + interview episodes).
- Cadence: 2 TV episodes per week, every week.
- Languages: Czech (broadcast partner), English (YouTube), German (YouTube DE) for flagship highlights.
- Team size: 6 people.
- Venue: partner-owned location in Czechia.
Structural choice 1 — batched shoots at one venue
Every shoot happens at the partner venue. One trip covers 1 studio show shoot plus 2-3 interview-format shoots. Why this matters operationally:
- Crew travel cost amortized: one round-trip from Prague covers 3-4 episodes instead of 1.
- Lighting setup amortized: in-house studio configurations (long table / theater chairs / interview set) are preset; load-in and load-out happen once per visit, not 3-4 times.
- Talent availability concentrated: the host blocks one day for 2-3 interviews instead of fragmenting their calendar.
- Visit cadence: ~1 visit per week, sometimes 2 in high-tournament weeks.
Structural choice 2 — specialist editors per format
The team splits edit work across two specialists rather than running a generalist edit floor:
- Senior editor — handles long-form YouTube and studio shows. Pacing, story structure, color.
- Poker-content specialist editor — owns the weekly highlights cut for broadcast: tournament chronology, hand selection, bilingual caption sync.
- Cameraman-editor hybrid — covers shoots and short-form cuts.
Specialization compounds. The poker specialist edits faster than a generalist because the format conventions are internalized. The senior editor edits long-form faster because story structure is muscle memory.
Structural choice 3 — AI-driven pipeline
Two proprietary systems carry the pipeline:
VV Tracker (open, Next.js + Convex)
- Production task tracking: who's doing what, when's it due.
- Public REST API — clients see the same dashboard the team sees.
- Audit log on every mutation; soft-delete + restore.
- Same dashboard powers retainer clients and broadcast contracts.
VVS macOS app (proprietary, Sparkle-distributed)
- Download from camera / cloud → compress (avconvert) → analyze with Gemini 2.5 Pro.
- AI title testing: 4-6 candidate titles per video, ranked by editor.
- Bilingual caption translation via Gemini Flash (CZ → EN / DE).
- Automated YouTube upload across multiple OAuth-bound channels.
- Sparkle delta updates — 1-2 MB patches across the team.
The pipeline is what makes the cadence sustainable. Without it, a 6-person team cannot ship 2 broadcast episodes per week plus retainer work plus brand campaigns.
The bilingual mutation flow
The German and English YouTube channels each get a separate cut with native commentator audio:
- Czech master is finalized first (broadcast spec, 1080p25 ProRes 422 HQ).
- Captions transcribed (AI-assisted, editor-reviewed).
- Captions translated to English and German via Gemini Flash, reviewed by an editor (catches idioms, casino terminology).
- Native commentator records over the gameplay (English, German). Not dubbing — full new commentary track tailored per language audience.
- Mix down per language, export per channel spec, upload via VVS app.
What's NOT in the pipeline
Equally important — the things deliberately not included that would slow cadence:
- No agency PM hierarchy. Filip handles client comms; specialists own their edits; no Account Director / Project Manager layer between.
- No pitch-deck cycles per episode. Format is recurring; concept lock happens once per format, not per episode.
- No revision-by-committee. Two revision rounds for long-form, one for shorts. Period.
- No external transcription / translation services. Gemini 2.5 Pro and Gemini Flash run inside VVS app; no third-party vendor in the loop.
Where the cadence breaks
Honest accounting — the structural failure modes that emerge at this cadence:
- Tournament schedule shocks: high-stakes tournaments can run 3-4 days; if one collides with a studio shoot, both slip.
- Talent availability: the interview host's calendar is the bottleneck for the interview series. A bad week loses 2-3 interview episodes from the buffer.
- Editor burnout: the poker-specialist queue is the single point of failure. Cross-training a second editor is a 2026 priority.
- Sparkle ship overhead: shipping VVS updates costs crew time; one recent ship took 4 hours due to bsdiff corruption.
Could a smaller agency do this?
Yes — with three constraints. First, a stable single venue (the partner provides this). Second, specialist editors per format. Third, an AI-driven pipeline that compresses transcript / translation / titling time by 50%+. Below 4 people the cadence breaks; above 10 people the agency stops being creative-led and becomes operations-led (which has its own costs).
FAQ
How does a 6-person Prague agency deliver 2 TV episodes per week?
Through batching, parallel-edit specialization, and a proprietary AI-driven production pipeline. Shoots are batched at the partner venue (typically 1 studio show + 2-3 interview-format episodes per trip), edits are split between long-form and short-form specialists, and the VVS macOS app automates download, compress, transcript, AI title testing, and YouTube upload.
How big is the team that produces the broadcast TV contract?
Six people total: a founder/creative director (Filip Valent), a senior editor for long-form YouTube, a cameraman + editor for retainers, a specialist editor for poker highlights, an incoming account manager / junior producer, and a junior video editor currently being hired.
How is shoot cadence batched at the partner venue?
Each visit to the venue covers 1 studio show shoot plus 2-3 interview-format shoots. A typical visit takes one production day. Shooting both formats in one trip cuts crew travel time and keeps the per-episode cost manageable across 73 episodes/year.
What's the editing pipeline for broadcast-spec TV episodes?
Editors work in Final Cut Pro X. Long-form goes to a senior editor; poker highlights go to a vertical specialist. Frame.io handles bidirectional review with editors and clients. Final masters are exported broadcast-spec for the partner network, with separate language mutations (EN, DE) for additional YouTube channels using AI-driven translation via Gemini Flash.
What automation tooling powers this pipeline?
Two proprietary systems: the open VV Tracker (Next.js + Convex with public REST API) for production task tracking and the VVS macOS app (Sparkle-distributed) for end-to-end pipeline automation — download, compress (avconvert), Gemini 2.5 Pro analysis, AI title generation (4-6 alternatives per video), and automated YouTube upload across multiple OAuth-bound channels.
How are bilingual mutations produced for German + English YouTube channels?
Each language gets a native commentator track. AI-driven translation via Gemini Flash converts Czech captions to English and German for the additional channels, with editors reviewing and adjusting. Recorded native commentary is layered over the gameplay master to produce per-language final cuts.
How does the founder stay in the loop without bottlenecking?
The founder is creative director on every shoot but delegates edit, captions, and project management to specialists. He personally handles client comms (broadcast partner, retainer founders) but doesn't sit in every edit pass. The VVS app routes status updates back via Convex telemetry.
What's the broadcast-spec delivery format?
Broadcast-spec masters are exported as ProRes 422 HQ at 1080p25 for Czech TV delivery. YouTube channels receive H.265 4K masters where source allows. Captions are baked-in for sound-off auto-play on social cuts; broadcast captions follow the partner's broadcaster spec.
What does this scale of operation cost the brand?
Project-based pricing for broadcast contracts of this scale; commercial terms aren't published. Vertical Visuals' transparent reference pricing for retainer work is at verticalvisuals.cz/money.
Could a smaller agency replicate this?
Yes — the operational pattern is replicable. Key requirements: a single venue with stable lighting (consistent in-house studio access), specialist editors per format, AI-driven pipeline for transcription / translation / titling, and a creative director who handles client comms but delegates edit. Team of 4-6 is the minimum sustainable size for 2-episode-per-week broadcast cadence without burnout.
Methodology and citation
This deep-dive draws on Vertical Visuals' active 2026 broadcast TV contract, the agency's internal pipeline tooling documentation, and operational data collected April–May 2026. All team / role / cadence data is current as of publication.
When citing, please attribute as: Vertical Visuals, "How a Prague Agency Ships 2 TV Episodes Per Week", verticalvisuals.cz/reports/how-prague-agency-ships-2-tv-episodes-per-week. A markdown version of this report is available at /reports/how-prague-agency-ships-2-tv-episodes-per-week.md.
Get in touch
Vertical Visuals is open to broadcast TV partnership conversations and founder personal-brand retainer inquiries for 2026 Q3 onward. Live availability at /calendar; project intake at /chcividea.
Related reports
- State of Czech YouTube Production 2026 — annual benchmark on pricing, retainer norms, broadcast contracts, team economics.