01 / 16
01 · Cover

Showspring

AI production studio for episodic podcasts and YouTube shows. Built around a show bible, not a prompt.

Coming soon · 2026
02 · The problem
The gap

AI tools make clips.
We make the next episode.

Clip tools Stateless generators. One prompt in, one clip out. Episode two doesn't know episode one existed.
Showspring Stateful studio. Characters, voices, formats, callbacks — all persistent. Episode 47 belongs next to episode 1.
03 · The insight
The insight

A show needs memory.

Clips don't have any. Every AI video tool on the market is amnesiac — it forgets your characters, your voice rules, your running jokes the moment you hit generate. That's fine for a one-off. It's fatal for a series.

04 · The primitive
The primitive

The show bible is a first-class object.

Every generation step reads it. Every episode writes back to it. A single source of truth for everything that makes your show your show.

Characters
Names, personalities, visual refs
Who's in the cast. How they speak. What they look like across every scene.
Voices
Per-character ElevenLabs IDs
Voice clones or voice picks, pinned to characters. No voice drift between episodes.
Formats
Segment rules & pacing
News-desk tempo, quest-narration pacing, sign-off lines. Structural DNA for each segment.
Canon
Running jokes & callbacks
The things you've established — pepperoni map, Kevin the Cat, the Gazette desk. Writer-room memory.
05 · Progressive, not upfront
How you start

You don't author a bible.
The bible writes itself.

Start with one character and one format. Make episode one. The bible fills in from what you actually make — voices that stuck, jokes that landed, segment rules that emerged. By episode two, the tool already knows your show.

06 · The outcome
The proof that it worked

Episode 47
sounds like episode 1.

Not because you wrote it the same way. Because the system remembered everything in between.

07 · Pipeline
The pipeline

Six steps. One workspace. One show bible through all of them.

01
Idea
Four AIs pitch episodes. You pick the best.
02
Script
Bible-aware dialogue. Characters stay in voice.
03
Voice
Per-character ElevenLabs, with scratchpad.
04
Visuals
Three engines, one consistent look.
05
Edit
Multi-track timeline. Audience reactions.
06
Publish
One click. Your channels, your metadata.

Each step reads and writes the bible. Every episode makes the next one smarter.

08 · Step 01
Step 01 · Idea

Four writers. One editor.
Pick the best pitch.

Claude, Gemini, Grok and a local Ollama model each pitch an episode concept in parallel — bible-aware from the first token. You see all four, pick one, kill the rest.

  • Bible-aware pitches — characters only act in-character, no canon violations.
  • Structural diversity — four models, four angles. No single-provider bias.
  • Grok verdict synthesis — optional final pass to compare and recommend.
Creative Director panel with five flow cards — Recommend Something, I Have an Idea, Research and Debate, Recurring Segment, Use Saved Idea
09 · Step 02
Script Writer panel — Grok 4 writes bible-aware dialogue with per-clip template, target length slider, structured clip blocks
Step 02 · Script

Dialogue that stays in character,
line after line.

Scripts inherit the show bible automatically. Recurring jokes, speech rhythms, voice consistency, segment-specific rules. Click any line to regenerate just that one.

  • Per-clip regeneration — touch one line without blowing up the scene.
  • Format-aware pacing — news-desk tempo vs. quest narration vs. interview rhythm.
  • Voice-gated generation — characters without assigned voices don't get dialogue.
10 · Step 03
Step 03 · Voice

Per-character ElevenLabs voices,
with controls.

Voice IDs live in the bible. Scratchpad previews every line before commit. Share-for-review links let collaborators sign off without touching the app.

  • Pinned voice per character — no drift between episodes, ever.
  • Scratchpad preview — play any line before it touches the main pipeline.
  • Share-for-review — private link, no login required.
Voice readout panel showing a clip with scene description, dialogue line, clip navigation, and per-character voice profiles
11 · Step 04
Scenes panel — grid of reference frames with character refs and dialogue previews, regenerate per clip
Step 04 · Visuals

Three video engines.
One consistent show.

Veo, WAN, and LTX each have strengths. The bible picks the right engine per clip, feeds it character references from the cast, and regenerates individual clips without disturbing the rest.

  • Engine routing — per-clip model selection based on scene type.
  • Character ref injection — every frame references the cast bible.
  • Per-clip regen — fix one shot, leave the rest alone.
12 · Step 05
Step 05 · Edit

Multi-track timeline,
ready to finish in DaVinci.

Trim, reorder, layer audience reactions, music beds and SFX. Export straight to YouTube, or hand off a properly-timecoded master to your NLE of choice.

  • Audience reaction track — 9 reaction types, SNL Weekend Update profile.
  • Lane-packed timeline — overlapping cues stack visually without collision.
  • DaVinci handoff — FCP7 XML or direct render with correct timecode and frame rate.
Trim timeline with video preview, duration ruler, video track with clip tiles, and audio lanes with intro and dialogue waveforms
13 · Step 06
Publish to YouTube panel with HLS preview, download and Drive buttons, brand-account connection, and AI-generated metadata
Step 06 · Publish

One button.
Your channels, your metadata.

AI-generated title options, description, tags, and thumbnail. Brand-account YouTube upload, cross-post to Discord, and the master stays yours.

  • Three title variants — pick, or regenerate.
  • Brand-account upload — YouTube, plus cross-post hooks.
  • Ships to your feed — not a hosted SaaS garden.
14 · Architecture
Under the hood

Right model, right step.

No single-provider bet. Each pipeline step routes to the model that's best at it — and falls back gracefully when a provider is down, slow, or pricing-hostile.

Reasoning
Claude Opus
Script writer, bible-aware synthesis, audience-reaction planner fallback.
Fast thought
Gemini 3 Pro / Flash
Multimodal audio-cue planner, cross-check on complex dialogue.
Irreverent
Grok 4
Idea pitcher, verdict synthesis across the four-AI debate.
Local GPU
Ollama qwen3:32b
Primary audience planner on our 5090 — zero-cost, zero-latency, fully offline.

Plus ElevenLabs for voice, Veo / WAN / LTX for video, SDXL / FLUX / Z-Image for stills.

15 · The proof
Showspring workspace with Creative Director panel
Currently producing

Meet
The Doodle Cast.

An AI-animated podcast hosted by two doodles — Rusty and Oreo — shipping regularly to YouTube. Characters, voices, running jokes, a news-desk segment, a full fan Discord. Every episode comes out of this pipeline.

thedoodlecast.com · @The_Doodle_Cast

16 · End
Showspring

Coming soon.
Reach out if it's interesting.

If you have something you want to make — a show, a podcast, a weekly segment — we'd love to hear about it.

Thank you.