The same episode, made two ways.
Fourteen things that differ once the work stops being one clip and starts being a channel. The last two rows are the ones where a browser tab wins.
A service’s own web appOne prompt, one clip, one tab per service.
Scena StudioThe whole episode in one queue, priced before it runs.
A service’s own web app
Scena Studio
How many services
One service per site: three tabs, three logins, three bills.
39 AI services and 80+ models (13 video, 12 image, 4 gateways, audio & custom REST) in one queue and one script sheet.
The unit of work
One prompt, one clip, and you sit through the session.
The episode: 46 shots queued from a sheet in one pass.
While you are away
The tab must stay open; a sleeping laptop ends the run.
It keeps going: threads, retries and rate limits are handled.
When something fails
You notice and retry by hand. A flat key stops everything.
Automatic retries, failover to the next account, no double bill.
What it costs
Credits bought up front, with the margin inside.
The service’s API price, with an estimate before you run.
Where your files live
On their servers, for as long as the plan lasts.
In your own folder. Nothing leaves the machine.
Your API keys
You have none; you rent access.
Yours, encrypted with Windows DPAPI, never sent to us.
Consistency in a scene
Describe the character again in every prompt, and hope.
A lock puts the same image and text into every prompt.
Consistency across episodes
Nothing carries over; every episode starts blank.
Reference sets carry the cast into the next series.
Catching drift
You find it in the edit, after paying for both shots.
Each clip is scored against its locks on your GPU first.
Platform shapes
Prompt again for each ratio, and pay again.
One shot comes out at 16:9, 9:16 and 1:1.
Running several projects
Browser history. No project, no folders.
Every episode is a project; open as many as you like.
Editing a shot
Some edit a shot from a written instruction.
Not an editor: it extends, upscales, lip-syncs and assembles.
Getting started
Open a tab and type. Works on a phone.
A Windows install and a sign-in; ten minutes to the first shot.
Where the web app is the better choice
Not every job needs a queue. If any of these describe you, open a tab and keep your money.
You want one clip today and do not have a series behind it.
You are not on Windows, or you work from a phone or a tablet.
You want a brand-new model the day it launches, before anyone wires it up.
You do not want to hold API keys or think about per-job cost at all.
What the plans allow, as of 2 November 2026
The cap, not the price, is the wall. For one episode — 46 shots plus retakes, about sixty renders — the entry plan on every site runs out in the first scene. Pay-per-shot has no plan to size and no credits to burn; where the API rate is above the plan’s, the queue picks the cheaper tier for the shots that can take it.
ServiceEntry plan givesPlan that fits 60 rendersPer clip on that planSame clip at the API rateCredits expire
Kling 3.x, 720p standard, 5 sStandard $10 → about 22 clipsPro (about 100 clips), $37/mo≈ $0.37≈ $0.28 — but API access starts at a $700 prepaid packagemonthly
Runway Gen-4.5, 5 sStandard $12–15 → 10 clipsMax (about 158 clips), $76–95/mo≈ $0.50not published in the sources openedStandard and Pro lapse monthly
Google Veo 3.1 quality, 8 sAI Pro £18.99 → 10 clipsUltra 5× (100 clips), £79.99/mo≈ $1.00$3.20 standard · $0.80 fast · $0.40 litemonthly and daily
Prices from the services’ own pages and the guides below, opened 2 Nov 2026; they move monthly, so treat them as the shape of the problem, not a quote. Open Test has no usage quota from Scena Studio. Provider prices apply to your own API keys.
Deep Cost & Workflow Audit · Desktop BYOK vs. Web Credit Wrappers
Why serious series creators move off web credit portals onto a desktop production pipeline
Web apps are built for casual single-clip prompting. When you produce a 46-shot weekly episode for YouTube or Meta Reels, web credit wrappers silently tax you 2.5x–4x on API usage, charge you for failed generations, lose character identity after shot 5, and leave you manually stitching 5-second clips in an external editor.
Side-by-side: Single-Model Web Apps vs. Web Credit Aggregators vs. Scena Studio PC
| Production Requirement | Single-Model Web Portal (Kling / Runway / Luma Web) | All-in-One Web Credit Wrapper (Higgsfield / Poe / Freepik) | Scena Studio Desktop PC (BYOK + Local GPU) |
|---|---|---|---|
| Cost for one 60-shot (300s) Episode | $45–$95 across 3 separate monthly subscriptions | $38–$72 in marked-up platform credits (2.5x–4x markup) | $4.80–$11.50 raw API price (0% markup) |
| Unused Monthly Credits | Expire at month-end or locked to one model | Expire on billing cycle reset | Never expire — you only pay per second rendered |
| Stitching 5–8s Short Clips into a Full Episode | Download 60 MP4s manually; jump cuts & face morphs | Manual download; no cross-shot frame handoff | Auto Last-Frame → First-Frame lock + FFmpeg crossfade stitch |
| Character Identity Lock Across 50 Episodes | Re-upload reference photos in every browser session | Session-bound; no local ArcFace drift gate | Persistent `{character:CODE}` + local GPU ArcFace ≥ 0.85 gate |
| Trademark (TM) & Copyright Pre-Flight Scan | None — you discover strikes after uploading to YouTube | None | Local TM logo, watermark & audio Content ID pre-flight shield |
| YouTube YPP & Meta Reels Monetization Proof | Raw unedited clips risk "Inauthentic Content" flag | No production audit trail | Exports Human-Direction Audit Log + 16:9 & 9:16 masters |