Nirvana — Creative Studio
Video, image, audio and episodic production — from script to final asset.
Four studios and thirteen squads covering the entire audiovisual chain. Script and treatment, image generation, video in Veo, Kling and Higgsfield, image-to-video animation, voiceover with cloned voice and post across six TTS providers. Episodic production and a reusable voice identity, from brief to final asset.
One-time payment. Per-buyer stamped download in the logged-in area.
What's inside
Businesses · 4
Produce institutional videos, advertising, pitch films and short-form content in which the client chooses the visual signature among acclaimed auteur-cinema canons — direction in schools like precision-noir, dialogue-driven…
Showrunner for content series. Produces, episode by episode, all the 360-degree deliverables of a channel (script, AI video, post, mini-PDF, ebook, landing page, creatives, ads), reusing characters, settings and voices via a bible…
A Nirvana business specialized in spectacular AI-first marketing/brand videos (Reels, ads, branded content 15-90s). The differentiator: a keyframe-first I2V pipeline with explicit vision QA (MCP nano-banana-pro describe_image) + overlays…
A business specialized in creating structured, flawless instructions for cloud-only TTS audio generation (Gemini 3.1 Flash TTS as default, with support for ElevenLabs v3, OpenAI gpt-4o-mini-tts, Hume Octave, Cartesia Sonic-3 and Azure Neural…
Squads · 13
Splits long source text into TTS-friendly chunks using the Murch boundary decision tree. Scores each candidate boundary (scene break 1.00 / paragraph 0.85 / sentence 0.70 / clause 0.40 / mid-clause forbidden).
Concatenates per-chunk WAV files via ffmpeg local with triangular crossfade (100ms default, 200-300ms at scene boundaries), then applies two-pass loudnorm to target LUFS (default -16, podcast standard), then exports to multi-format (WAV…
Calls cloud TTS provider APIs (Gemini default + fallback chain ElevenLabs/OpenAI/Hume/Cartesia/Azure) to render each chunk into a WAV file. Implements automatic fallback on 429/5xx/timeout. Resamples on provider mismatch.
Brand-consistent visual deliverables: extracts design systems from URLs (Refero + live extraction), generates PDFs, PPTX, social posts, carousels and programmatic videos via two paths (Veo 3.1 + Remotion for AI footage; HyperFrames…
Most capable at generating media via the official Higgsfield CLI (@higgsfield/cli), headless and agent-native: photo-realistic image (Soul 2.0 / Nano Banana), multi-model video (Kling 3.0, Veo 3.1, Seedance 2.0, DoP) with 50+ motion presets…
Generates photorealistic images and spectacular photos with gpt-image-2 via Codex: directed photography (light, lens, camera, composition, studio vs.
Top-grade editorial infographics: sharp narrative, grounded data, 2026 art direction and a quality gate.
Transforms a voice-seed.json + chunks_plan.json into provider-specific prompts (Gemini Layer Cake / ElevenLabs inline tags / OpenAI instructions / Hume voice_prompt / Cartesia SSML / Azure SSML).
Cross-engine routing and planning layer for AI video — understands the user's request, performs tool/cost arbitrage across 19+ engines (Veo 3.1, Sora 2, Kling, Runway, Luma, Wan 2.2, HunyuanVideo, LTX, Sync.so, LatentSync…
Reads a free-form user brief + source text and emits a structured brief-spec.yaml — extracts language, register, audience, deliverable type, provider override (if any), accent hints, gender/age hints, expected duration, and the first 500…
Specialist in Google Veo 3.1 image-to-video via GenAI: animates images with motion instructions, locks the image and animates only selected elements (cinemagraph), generates perfect loops, videos with multilingual speech and lip-sync, and series with…
Keyframe-first I2V for marketing reels and ads (15-90s). Decomposes brief → shot list → keyframes (Nano Banana Pro) → vision QA (MCP nano-banana-pro__describe_image) → I2V (Veo 3.1, applying the golden rule) → audio → Remotion overlays →…
Designs the canonical voice identity (voice-seed.json) from a brief-spec. Co-grounding in Andrea Romano (performance) + Geoff Lindsey (phonetics). Always emits a seed with cross-provider mapping pre-computed for all 6 supported providers.
How to install
- Install the engine:
npx @nirvana-os/cli - After purchase, download your stamped pack from the logged-in area and run
bun setup.ts - Update whenever you want:
nrv update creative-studio
Honest note
The squads and businesses generate real strategy, documents, code, copy, plans and reports on the Nirvana-OS engine. Image and video generation uses the tools in your environment; publishing and execution on external platforms depend on your keys and integrations. The content is yours to use and adapt.