Everything between the script sheet and the finished episode.

Seven groups of work, each one a screen you can open from the left rail. Where a capability lives is printed beside it, so this page and the app use the same names.

Make the video

Six ways in, 13 active video services (35 models) plus 4 gateways & custom REST endpoints behind them.

VEO 3.1 · 1080P Text → video One prompt, one shot. The settings ride on each row. Video queue

I2V · FACE LOCK Image → video A still starts moving, and the face stays put. Video queue

KEYFRAME A → B First–last frame Set the first and last frame; the service fills the middle. Video queue

MULTI-REF · 4X Reference mode Sheets, outfits and sets go in as pictures, not words. Video queue

EXTEND +5.0 SEC Extend More seconds on a good take. Nothing re-rendered. Upscale & extend

V2V RESTYLE Video → video Same motion, new look. Video queue

Make the images

12 image services (47 models) and a local Real-ESRGAN upscaler.

12 IMAGE SERVICES · 47 MODELS Batch image generation The video queue’s engine, for stills. Image queue

TURNAROUND LOCK Character sheets and props Make the reference set, then lock it. Reference sets

SCENE & CODE FILTER Image library Every image of the episode, filtered by scene or code. Image library

VULKAN 4× ESRGAN Upscale on your own GPU 2× to 8× on your own GPU. No per-image fee. Upscale (local)

Keep it the same show

The part a web app has no answer for.

CHAR_NOAH · LOCKED Reference locks A code like NOAH stands for one image and one line. Reference sets

PRE-FLIGHT BLOCK Continuity rules A BLOCK stops a job before it costs money. Continuity rules

RESOLVED TOKENS Prompt inspector See where every part of the final prompt came from. Continuity rules

GPU + VISION AI 98.4% Consistency check Face, outfit & palette scored on GPU + Vision AI. Consistency check

FROZEN TAKE ✓ Approvals and freeze An approved take cannot be overwritten. Approvals & locks

Run it at scale

Forty-six shots is a normal evening.

SQLITE CRASH-SAFE Resumable queue Survives a crash. Never bills twice. Video queue

4 THREADS · AUTO-RETRY Threads and retries Jobs run side by side and retry on their own. Settings

KEY POOL FAILOVER Account failover An empty key hands over to the next account. Accounts

HOT-SWAP MODEL Change service mid-queue Move jobs to another model, settings intact. Change service

LIVE QUEUE MATRIX Queue manager Reorder, pause and re-prioritise. Queue manager

Know what it costs

Before you spend it, not after.

PRE-RUN $18.40 EST Queue estimate The whole run’s price, before you press RUN NOW. Video queue

$0.38 / SHOT LOGGED Per-job cost What each finished job really cost. Cost tracking

LIVE WALLET HUD Session and balance Session total and key balances, always on screen. Cost tracking

DPAPI · 0% MARKUP Your keys, their prices Your keys, at the services’ own prices. No markup. API services

Run many projects

A channel is a hundred videos, not one.

EE012 PROJECT TREE Episode as a project Its own code, folder, sheet and render tree. New episode

SHARED SERIES CAST Cast across series The same cast carries into the next series. Reference sets

XLSX DIFF RE-IMPORT Script sheet as the source Plan in Excel. A re-import sends only changed rows. Script sheet

{SCENE}_{SHOT}_v2 Naming pattern Rename a whole episode by editing one field. Settings

Finish and hand over

Out of the app and into the edit.

16:9 · 9:16 · 1:1 Every aspect ratio 16:9, 9:16 and 1:1 from one shot. Export

ELEVENLABS + SYNC Dialogue and lip-sync A voice per line, timed against the shot. Dialogue & voice

4K MP4 · XML · SRT Export the episode Join the episode with audio and subtitles. Export

LOCALHOST WEBHOOK Webhook and API Other tools push jobs in and read status out. Webhook & API

What it deliberately does not do

A tool that claims everything is a tool you cannot plan around. These are the jobs Scena Studio hands back to you, on purpose.

No timeline, no trimming, no cutting — your editor keeps that job. No object removal, rotoscoping or masking. No multi-track music scoring or DAW mixing — dialogue is generated, speed-fitted and stitched per shot. No hosting, no publishing and no scheduling to YouTube, TikTok or Meta.

Complete AI Service & Model Catalog in Build 1.1.0 (39 Services · 80+ Models)

SYNCED WITH DESKTOP CATALOG.CS

Connect direct vendor API keys, route through a single unified gateway key, or attach your own custom REST/OpenAI-compatible endpoint via API Services → + Attach Custom AI Provider.

13 Active Video Services (31 Generation + 4 V2V Models) Kling (4.0, Omni O3, 3.0 Pro/Std) · Google Vertex AI (Veo 3.1, 3.1 Fast, 3.1 Lite) · Google Gemini API (Gemini Omni 1.1 Flash) · Runway (Gen-4.5, Aleph 2.0 video edit, Gen-4 Turbo) · Luma (Ray 3.2, Ray 3 HDR/EXR) · MiniMax / Hailuo (H3 stereo, Hailuo 2.3/Fast) · Seedance (2.5 up to 30s & 30 refs, 2.0/Fast) · Alibaba Wan (3.0, 2.7, 2.6) · Alibaba HappyHorse (1.1, 1.0) · xAI Grok Imagine (Video 1.5) · Vidu (Q4, Q3, Q2) · Pika (2.5) · LTX Lightricks (LTX-2.5 Pro 4K, 2.5 Fast, LTX-2) · plus Sora 2 Pro & Sora 2 via fal.ai.

12 Image Services (47 Models) + Local GPU Upscaler GPT Image (2, 1.5, 1, DALL·E 3 HD) · Nano Banana (2, Pro Gemini 3, Gemini 2.5/2.0 Flash Image) · Imagen 4 (Ultra, 4, Fast, 3.0) · Grok Image (Imagine 2.0, 1.5, Aurora) · Seedream (4.5, 4.0 Pro, SeedEdit 3.0 Character Edit) · Flux 2 & Specialist Suite (Flux 2 Pro/Dev, 1.1 Ultra 4MP, Kontext Max/Pro Lock Char, Fill Pro Inpaint, Kolors 2.1 & Virtual Try-On Outfit, Qwen-Image 2.0/Edit, Hunyuan 3.0, Hidream I1/E1, Wan 2.6 Image, Runway Gen-4 Image) · Ideogram (v3 Quality/Turbo, Character Lock) · Recraft v3 (Vector/Raw/SVG) · Midjourney (v7, v6.1, Niji v6) · Stability AI (SD 3.5 Large/Turbo/Medium, Ultra) · Leonardo.Ai (Phoenix 1.0, Kino XL, Anime XL) · Luma Photon (Photon 1 Lock Char, Flash).

Voice TTS, Speech ASR, Translation & V2V Lip-Sync / Upscale Voice TTS: ElevenLabs (v3 with emotional tags, Multilingual v2, Flash v2.5), OpenAI TTS (Realtime mini WebSocket, gpt-4o-mini-tts, tts-1-hd), Azure Neural HD (SSML) · Speech ASR: Whisper large-v3 (local GPU $0), OpenAI Whisper-1, Deepgram Nova-2/3, Gemini 2.5 Flash · Translate: DeepL, Google v2/v3, GPT-5.1, Claude Sonnet 4.5 · V2V Lip-Sync & Upscale: Kling Lip-sync, sync-lipsync 2 (4K), Hedra Character 3, Topaz Video Upscale (2×–4×).

4 Unified Gateways, Vision AI Inspector & Custom REST Endpoints Gateways: fal.ai, WaveSpeed, Replicate, OpenRouter · Hybrid Consistency Inspector: Local GPU CIE76 ΔE palette + 64-bit dHash + portrait face likeness ($0) paired with optional Cloud Vision AI (Gemini 2.5 Flash, Claude Sonnet 4.5, GPT-4o-mini) · Custom Providers: Attach any OpenAI-compatible, fal queue, Replicate, or custom REST JSON endpoint with Bearer / x-api-key auth and Polling or Webhook callbacks.

Built for Monetizable Channels · High-Demand Production Modules

Four production bottlenecks that kill AI channels — and how Scena Studio solves them inside the app

Generating a cool 5-second clip is easy. Turning 60 short clips into a cohesive, copyright-safe, monetizable 8-minute episode every week is where creators get stuck. Here are the actual screens and engines built into Scena Studio for Windows:

Short clip auto-stitching with Last-Frame to First-Frame character lock

01 · SEAMLESS SHORT-CLIP AUTO-STITCHER

Stitch 5–8s clips into a continuous episode with zero character drift

Automatically extracts the final frame of Shot N (`frame_last.png`) to anchor Shot N+1 while keeping `{character:CODE}` and `{location:CODE}` locked. Local FFmpeg normalizes frame rates to 24fps and applies `-3dB` audio crossfades so viewers never feel a 5-second seam.

Pre-Flight Trademark and Copyright Shield Scanner

02 · PRE-FLIGHT TM & COPYRIGHT SHIELD

Catch trademarked logos, watermarks & Content ID risks before upload

AI models occasionally hallucinate brand logos on shirts, soda cans, or street signs, or leave subtle model watermarks. Scena Studio scans every rendered frame on your GPU for trademark/watermark artifacts and verifies audio stems, plus exports a Human-Direction Audit Log for YouTube YPP & Meta reviews.

Multi-ratio 16:9 YouTube and 9:16 Meta Reels export with smart tracking

03 · DUAL-REVENUE 16:9 + 9:16 SMART RE-FRAMING

One episode project → YouTube Long-Form + Meta Reels & TikTok

Instead of paying twice to re-render vertical clips, Scena Studio uses local YOLO/ArcFace face tracking to pan-and-scan your 16:9 4K master into perfectly framed 9:16 vertical shorts with safe-zone subtitles for Meta Reels Bonus and YouTube Shorts.

Resumable overnight SQLite queue and BYOK cost estimator

04 · OVERNIGHT SQLITE QUEUE & ZERO-MARKUP BYOK

Queue 60 shots overnight at raw API rates with automatic key failover

See the exact dollar invoice before clicking Run. If a provider hits HTTP 429 rate limits or a key runs low at 3 AM, the SQLite WAL queue rotates to your backup key without dropping a single frame or double-billing.