# PV30

> PV30 is an AI video agent. Chat with a producer assistant to write the script, design every shot, generate consistent images, video and voice, and render the final cut.

Canonical: https://www.pv30.com/

## What is PV30?

PV30 is a web-based AI video agent. A conversational producer assistant takes a project from idea to final cut — writing the script, breaking it into characters, environments, props and storyboard shots, generating images, video clips and voice-over, and rendering the timeline — while every result stays editable on a visual node canvas. It ships with three video types: ads, short dramas, and tutorials & explainers.

- **Category:** AI video agent (web app)
- **Video types:** Ad video · Short drama · Tutorial & explainer
- **Starting points:** Creation details · your own script · a reference video
- **Pipeline:** Script → elements & shots → images → video & voice → timeline → final cut
- **Control:** Chat, node canvas, versioned assets, review & revise at every stage
- **Spending:** Credit quotes confirmed by you, or Auto mode within a budget you set

## One agent. Three kinds of film.

The video type sets the creation form, prompt templates and default skills. Pick one when you create a project.

- **Ad video** (10–120 s · default 30 s) — From product name and selling points to a campaign-ready short ad. Inputs: Product, Selling points, Audience, Duration, Style.
- **Short drama** (30–180 s · default 60 s) — From a story idea and characters to a fast-paced, high-conflict drama script and film. Inputs: Story idea, Characters, Opening hook, Duration, Genre.
- **Tutorial & explainer** (30–300 s · default 90 s) — From a topic and outline to a well-structured explainer with narration. Inputs: Topic, Outline, Audience, Duration.
- **Visual styles:** Realistic, Anime, Chinese guofeng, Japanese manga, American comic
- **Film language:** Chinese or English audio & subtitles, with an optional story locale for culturally fitting characters and sets

## Never start from a blank page.

Three ways in — each one lands in the same editable project.

- **Fill in the creation details** — Give the essentials for your video type. The assistant writes the script from scratch using that type’s template.
- **Paste your script** — Imported exactly as written — no rewriting, translating or reordering. PV30 then parses it into shots and elements.
- **Upload a reference video** — PV30 transcribes the audio, analyses key frames and assembles a script that follows its rhythm and structure.

## From idea to final cut, in five stages.

Run it end to end with one-click film, or take it one step at a time. Each stage saves real, editable results.

1. **Script** — Create from details, import your text verbatim, or extract from a reference video. Revise the whole script in place whenever you like.
2. **Elements & shots** — The script is parsed into characters, environments and props, and into storyboard shots that reference them. Shot numbers stay stable as you reorder.
3. **Storyboard design** — Each shot gets beats, framing, camera angle and movement, performance, lighting, keyframes, lines and duration. Review it, then revise from the notes.
4. **Images, video & voice** — Generate element images and storyboard frames, then shot videos and voice-over. Preview the exact prompts for free before you pay for video.
5. **Timeline & delivery** — Arrange, trim and swap clips, set transitions and audio tracks, render a preview, then the final cut. Download the film or every asset.

_Generation depends on the image, video and voice providers configured for your workspace._

## An agent you can direct.

- **A producer, not a prompt box** — Streams its thinking, tool calls and step-by-step progress, and suggests next actions you can send with one click.
- **Queue, stop, resume** — Send follow-up instructions while it works — they queue in order. Stop the active turn, or reconnect to a running session later.
- **Quotes before you pay** — Paid steps come with a server-side credit quote you confirm. A promise in chat is never enough to spend.
- **Auto mode with a budget** — Turn on Auto and the assistant approves allowed quotes within your session budget. Over budget or destructive? It asks you.
- **Shot-level direction** — Beats, framing, camera moves and lighting per shot, plus a structured review-and-revise loop that snapshots before it changes anything.
- **Consistent faces and places** — Characters, environments and props have reference images reused across shots. Any shot can start from a chosen last frame.
- **Reference-video replication** — Borrow a clip’s motion, camera movement or effects for a shot, anchored to your own character images. Not face swap, not a copy.
- **Skills and visual styles** — Directing, editing and prompt-writing skills matched to each model, plus five visual styles from realistic to anime and comic.
- **Versions you choose** — Regenerating never overwrites. Each node binds the exact asset version you pick; upstream changes mark dependents stale.
- **Real timeline, real render** — Clip order, trims, transitions, source audio and separate voice-over tracks — rendered to a preview and a final cut you can download.
- **Long jobs that survive** — Video jobs persist in the background. Close the tab; the Task Center tracks every job across projects.
- **Built for teams** — Company accounts with members and credit allocation, a credit ledger and price list, and an admin console for models, skills and video types.

## Hand the work to AI. Keep the decisions.

Everything the assistant makes lands on a node canvas you can inspect and change.

- **Pick the take you like** — Compare versions of a character, environment or shot and bind the one that fits. Other nodes keep their own choice.
- **See what changed** — Switch an upstream version and dependent nodes are flagged stale — kept as-is until you decide to regenerate.
- **A real asset library** — Resumable uploads for images, video and audio, version history, references in chat, a trash you can restore from, and copying versions across projects.
- **Focused node detail** — Double-click any node to open it with its directly related nodes, its bound inputs and the exact generation inputs behind each version.

## Pick the model per job.

PV30 orchestrates several providers and keeps each model’s real capabilities — it does not pretend they are the same.

| Kind | Models | Used for |
|---|---|---|
| Images | Seedream 5.0 Lite (default) · Seedream 5.0 Pro · FLUX | Character, environment and prop images; storyboard frames |
| Video | Seedance 2.0 Mini (default) · Seedance 2.0 Pro | Multi-image reference, first frame, one-click film, reference-video replication |
| Video | MiniMax H3 | Text, first / last / first-and-last frame, up to 9 references; 768P or 2K; 4–15 s per clip; durable jobs |
| Video | LTX-2 | First-frame generation on a self-hosted service |
| Voice | MiniMax TTS | Voice-over tracks |

_Availability depends on provider accounts and workspace configuration. Generation tier and final render resolution are set separately._

## For teams that need more video, not more headcount.

- **Brands & e-commerce** — Turn selling points into campaign ads, and test more creative angles while characters stay on-brand.
- **Short-drama studios** — Go from story idea and characters to hooked, shot-designed episodes with consistent cast and sets.
- **Educators & explainers** — Turn a topic and outline into a structured explainer with narration.
- **Agencies & creators** — Remix a reference video’s structure or camera language into original work, with reviewable storyboards.

## PV30 vs. a single text-to-video prompt

|   | PV30 | Single-prompt text-to-video |
|---|---|---|
| Starting point | Creation details, your script or a reference video | One text prompt |
| Output | Multi-shot film with timeline, transitions and voice-over | One short clip |
| Direction | Per-shot framing, camera, performance and lighting | Whatever the prompt implies |
| Consistency | Shared element references and last-frame chaining | Re-described each time; drift between clips |
| Editing | Change one line, shot or asset and regenerate only that | Rewrite the prompt and start over |
| Cost control | Confirmed quotes or a capped Auto budget | Varies by tool |

## Questions, answered.

### What is PV30?

PV30 is an AI video agent. A producer assistant takes your project from script to storyboard, images, video, voice-over and a rendered final cut, and you can edit every result on a visual node canvas.

### What kinds of video can I make?

Three video types are built in: ad videos (10–120 s), short dramas (30–180 s) and tutorials & explainers (30–300 s). Each has its own creation form, prompt templates and default skills, and administrators can edit or disable types.

### Do I need a finished script?

No. Fill in the creation details for your video type and the assistant writes the script. If you already have a script, paste it and PV30 imports it unchanged; you can also start from a reference video.

### How does PV30 keep characters and locations consistent?

Characters, environments and props are key elements with their own reference images, reused by every shot that uses them. A shot can also start from the exact version of a chosen last-frame image.

### Can I copy a video I like?

You can extract a script from a reference video, and with Seedance models you can replicate a clip’s motion, camera movement or effects for a shot, anchored to your own characters. It is not face swap, lip-sync or a one-to-one copy — make sure your use respects third-party rights.

### Can I edit after generating?

Yes. Edit lines, shots and elements, add, remove or reorder shots, compare and bind asset versions, and regenerate only what needs it. Text edits never silently regenerate paid media, and nodes affected by an upstream change are marked stale.

### Which AI models does PV30 use?

Images: Seedream 5.0 Lite / Pro and FLUX. Video: Seedance 2.0 Mini / Pro, MiniMax H3 and LTX-2. Voice: MiniMax TTS. Which ones are available depends on how your workspace is configured.

### How is usage billed, and what is Auto mode?

PV30 uses credits priced by model and resolution. Paid steps show a quote you confirm. If you switch on Auto mode, the assistant approves allowed quotes within your session budget and asks you again when it would exceed the budget or do something destructive.

### What happens if I close the browser during generation?

Video jobs run in the background and appear in the Task Center. One-click film keeps reporting progress in the assistant conversation; stopping it prevents further steps but does not cancel jobs already submitted to a provider.

### Which languages are supported?

Films can be made with Chinese or English audio and subtitles, with an optional story locale. The PV30 app interface is in Chinese and English; this website is in English, Chinese, Japanese and Korean.

---

Screenshots and frames on this page come from demo projects made in PV30; brands and characters are fictional. Results depend on the models and settings used.