Skip to content

What Onira does. Where it stops.

The current product boundary for documentary production, review, duration, models, source handling, repair, rights, and publication. Last reviewed July 11, 2026.

Current product

A documentary production workflow, not a universal video editor.

Onira coordinates a brief through story development, measured narration or dialogue, screenplay assembly, authoritative subject and setting references, scene imagery, motion, music, subtitles, timeline construction, and final rendering.

The strongest current fit is original, narration-led history, civilization, and cultural-heritage video. Other formats may work, but they are not the initial public promise until evidence and workflow controls support them.

Input
A production brief with subject, audience, duration target, tone, language, aspect ratio, source boundary, and review expectations.
Story
Classification, internal research assistance, showrunner development, story blueprint, audio script, outline, and assembled screenplay.
Audio first
Narration and dialogue are generated and measured before final visual architecture so shots conform to the accepted voice track.
Visual production
Reference-aware scene images, motion generations, continuity memory, music, subtitles, and a canonical Remotion timeline.
Output
A final MP4 for creator review and manual publication. Rendering is a separate lifecycle after pipeline completion.

Models and routing

The product owns the production route; models remain replaceable departments.

Language work
OpenAI GPT-5.4 for the active LLM routes.
Scene images
Gemini 3.1 Flash Image through AI Studio for the current image pipeline.
Scene motion
Pixverse v6 for the normal profile and Veo 3.1 Fast for the high profile.
Speech and music
ElevenLabs eleven_v3 for speech and ElevenLabs Music for score.
Assembly
Remotion for the final timeline and MP4 render.

Model access is not a guarantee of a specific visual result. Routes can change as capability, reliability, cost, and provider contracts change; dated technical pages should be refreshed when they do.

Duration and plans

Plan ceilings are not quality guarantees.

Plan limits
Creator supports selectable targets up to 10 minutes, Studio up to 20, and Pro or Enterprise up to 30.
Available targets
The current app picker is authoritative for the targets available to a workspace under its active plan.
Quality boundary
A plan limit is not a guarantee of long-form quality, fixed cost, turnaround time, audience retention, or publishing readiness.
Production guidance
Choose the shortest target that can sustain the evidence and story, inspect the expected-to-maximum credit estimate, and review the finished film.
Founding cohort
The cohort begins with bounded 1-, 2-, or 3-minute paid proofs as an editorial service policy, not a separate engine limit.

Credits are usage units, not video minutes. The application shows an expected-to-maximum preflight estimate, reserves the maximum, caps billing at that amount, and settles against completed operations.

Review and repair

Durable checkpoints reduce rework; they do not eliminate editorial review.

Blueprint review
REVIEW can pause the factual workflow for story approval before expensive downstream work. AUTO also exists, so public copy must not imply universal human approval.
Resume
Eligible user resumes start from the earliest failed checkpoint, delete that step and downstream checkpoints, and preserve completed parallel siblings.
Director chat
When deployed, chat is limited to regenerating one selected PREVIEW video clip. Other Studio actions are separate direct controls.
Accounting
Provider-call records support observability; pipeline operation logs remain the billing source of truth.

Factual boundary

Research-assisted; creator verification required.

The internal research stage can improve the factual starting point, but the current customer product does not guarantee grounded research for every run or map every published claim to a visible source receipt.

Until a user-facing claim-to-source pack, confidence state, and creator sign-off pass end to end, Onira must not claim automatic factual accuracy or source traceability.

Creator responsibility
Verify claims, dates, quotations, pronunciations, uncertainty, sources, rights, and the final publication package.
Reconstruction
Generated historical scenes must not be presented as archival evidence. Realistic altered media may require platform disclosure.
Sensitive subjects
Living people, current events, health, finance, conflict, crime, and contested claims require stronger sourcing and often expert or legal review.

Not shipped

Capabilities Onira does not currently promise.

Publishing
No YouTube or social upload, scheduling, channel management, title publishing, or autoposting.
Source receipts
No customer-facing claim-to-source mapping, confidence ledger, or verified source-pack export yet.
Perfect consistency
No guarantee of identical faces, wardrobe, objects, actions, geography, or physics across every generation.
General editing
No promise that every scene, pace, narrator, or film-level change can be directed conversationally.
Enterprise controls
No standard SAML SSO, audit-log product, white-label workspace, SLA, or multi-approver institutional workflow.
Creative controls
No voice cloning, public API, LUT or manual color-grading system, custom brand presets, or exact chart and UI rendering.
Outcomes
No guarantee of monetization, views, revenue, factual accuracy, legal clearance, advertiser suitability, or publication acceptance.

Rights and privacy

Output assignment has real legal boundaries.

Output
Generated output is assigned under the Terms, subject to applicable law, third-party rights, source material, and provider terms.
Customer material
Creators must have permission to use prompts, uploads, likenesses, trademarks, music, archive, and other source material.
Training
Onira does not use customer prompts or projects to train its own AI models. Production data is sent to providers as required to generate the requested work.
Publication
The creator decides whether a reviewed output is suitable to publish and remains responsible for the resulting use.

A narrower promise is a more useful promise.

Evaluate Onira on a complete documentary, the review work it requires, and the cost per accepted film, not on a selected five-second generation.