vivago R1 makes the asset library the real video workflow

vivago R1 turns a natural-language request into a video workflow spanning story, script, storyboard, characters, shots, and final generation. According to ifanr, its Chat interface handles intent and asset calls, while a canvas keeps scripts, key assets, storyboards, and videos visible in structured form.

That is a meaningful shift from prompting isolated clips. In a cyberpunk Journey to the West test, ifanr reports that vivago R1 chose Cinematic Story Director, produced a storyboard script and key images, then continued to video after confirmation. Yet the durable work remains in the pipeline: woshipm identifies scriptwriting, character locking, storyboard-image generation, image-to-video conversion, voice acting, and editing. The real question is how those assets, formats, and audience signals are controlled before a creator treats generated output as a finished production.

The agent starts a production, but does not approve it

vivago R1 places a familiar production chain inside a conversational interface. According to ifanr, a user can describe what they want in natural language, then move from story and script into storyboards, characters, shots, and final video generation. Chat interprets the request and handles asset calls. It also generates scripts, storyboards, and videos, while a canvas keeps the working material visible in a structured format.

The conversation starts production; it does not settle every creative decision.

ifanr's cyberpunk Journey to the West short-film test shows the distinction. vivago R1 automatically chose Cinematic Story Director, produced a complete storyboard script, and created key images. Video generation continued only after confirmation. That pause matters because the storyboard and images are not merely intermediate outputs: they are the point at which a creator can judge what the agent has interpreted before committing the project to moving footage.

The underlying workflow remains staged. As woshipm describes it, AI comic drama begins with scriptwriting and character locking, then proceeds through storyboard-image generation and image-to-video conversion. Voice acting follows, and editing closes the process. vivago R1 can gather these tasks into Chat and the canvas, reducing the need to move between separate production surfaces. But its own confirmation step illustrates the limit of automation. The agent can call assets and prepare a proposed sequence; a person still needs to approve the direction before generation carries on.

Character locking turns generation into a reusable system

A usable continuity layer begins before video generation: a portrait, product packaging, and a logo become reference material that can be carried through the work instead of re-described from scratch. According to ifanr, Gouda's social-media advertising test used those inputs with Story Ad Director to generate an advertising storyboard, then produced the video after the plan was confirmed. The confirmation step matters. It separates approved visual material and a chosen plan from whatever the model happens to generate next.

That is how an asset library becomes part of the workflow, rather than a folder beside it.

The pressure grows quickly in comic-drama production. woshipm says a three-minute short episode needs about 60-80 storyboard shots, while a five-minute episode needs about 80-120 shots. At that scale, a character that shifts appearance from shot to shot is not a minor defect; it creates repeated correction work across the episode. A character reference image, plus character-reference or character-consistency settings, gives production a practical control point.

woshipm's author says settings weighted at 0.7-0.8 can reduce rework by 80% when used with a character reference image. That claim does not mean the model has solved continuity on its own. It means the reusable reference has become the constraint the model works against. The same logic applies to scenes and products: approve the material that must remain stable, confirm the storyboard, then generate within those decisions. Generation is still the visible step, but asset approval is what makes repeated generation manageable.

Short episodes fit the models better than a long drama

vivago R1's claim is consequential because it shifts the unit of generation from a clip toward an episode. According to ifanr, the conversational creation agent can stably generate video lasting 3 to 5 minutes. That duration matches a short dramatic installment more closely than a fragment designed to be edited into something larger. Stability at that length could reduce the assembly burden, but it does not make format planning optional.

Shot count is where the difference becomes visible.

According to woshipm, a 3-minute short episode needs about 60-80 storyboard shots, while a 5-minute episode needs about 80-120. Those figures describe the practical scale of an episode even before generation choices are made. A creator working from discrete shots must decide which moments deserve their own image-to-video pass, then preserve enough continuity for the assembled result to read as one story rather than adjacent clips.

Seedance 2.0 points toward the opposite production shape. woshipm says its image-to-video mode recommends 3-5-second videos at 720P or 1080P. That makes it suited to a shot-based pipeline, where the episode is built from many brief generations rather than expected from one sustained run.

For beginners, woshipm recommends a 1- to 2-minute trial work with about 30-40 storyboard shots. The smaller format exposes pacing and continuity problems early. It also makes short episodes a useful testing ground: creators can learn which moments hold together before attempting a 3- to 5-minute narrative.

A broadcast milestone is not proof of narrative quality

Mango's satellite-television drama marks a meaningful production milestone, but it does not settle the separate question of storytelling quality. According to a report, *Journey to the West Again* launched on Mango TV on August 31 and entered the prime-time slot of Hunan Satellite TV. The report describes it as China's first AI long-form drama broadcast on satellite television. That placement shows that an AI-made programme could move beyond a demonstration and into a high-visibility broadcast slot.

The production method is central to what that milestone proves. The report states that AIGC created the drama's visuals and character performances, with no live actors involved in production. Mango Lingchuang's AIGC platform supplied the end-to-end intelligent solution. Together, those details point to a complete production loop: the system supported the work from generation through a finished programme that Mango TV could distribute and Hunan Satellite TV could schedule in prime time.

Prime-time exposure proves delivery and attention, not dramatic control.

Long-form drama creates a different standard. According to woshipm, the Huaguo Mountain arc, the first season of *The Post-Journey to the West*, contains five episodes of 40 minutes each. A format of that length leaves more room for a narrative to reveal weak story design or uneven pacing. It has to sustain its material across episodes, rather than simply make an immediate visual impression.

That distinction appears in viewer response reported by woshipm. A netizen found *The Post-Journey to the West* visually striking, while identifying serious shortcomings in story design and narrative pacing. Visual generation and generated character performance can therefore be real production achievements without guaranteeing that a drama's episodes cohere. The Mango broadcast demonstrates an end-to-end AIGC production capability and an ability to reach a major screen. It does not, by itself, demonstrate that long-form narrative quality has been solved.

AI video production approaches: workflow control, format limits, and monetization

Build a controlled short-form pipeline before attempting long-form AI video

• You need recurring characters, products, or locations to remain recognizable across multiple shots or episodes. Create and approve reusable character and scene assets before generating at scale. Gouda stores generated, selected, and approved content in an asset library for reuse, while the AI comic-drama workflow explicitly includes character locking. One source claims that using a character reference image and character-consistency settings at 0.7-0.8 can reduce rework by 80%.
• You are a solo creator testing an idea for the first time. Start with a one- to two-minute trial rather than a long episode. The source recommends roughly 30-40 storyboard shots for such a test; a three-minute episode needs about 60-80 shots, and a five-minute episode about 80-120 shots.
• Your plan relies on image-to-video generation for many individual shots. Design the edit around short clips and assembly, not a single uninterrupted generation. Seedance 2.0 recommends 3-5-second videos at 720P or 1080P, and the documented AI comic-drama process moves from storyboard images to image-to-video, voice acting, and editing.
• You have a broad creative brief but need a reviewable production process. Use an agent for script and storyboard development, then approve the plan before video generation. In documented tests, vivago R1 generated a storyboard script and key images before generation continued after confirmation; Gouda similarly generated an advertising storyboard before producing video.
• You are deciding whether to charge viewers immediately for AI-made episodes. Prioritize audience testing and trust before monetization. In one survey, 74.68% of respondents had watched AI short dramas, but 86.52% had never paid; among those who had paid, 83.87% had spent less than 50 yuan in total.

Views and paywalls need validation before they become a business

Views establish reach, not a payment habit. woshipm reports that 74.68% of survey respondents had watched AI short dramas, yet 86.52% had never paid for them. Among respondents who had paid, 83.87% had spent less than 50 yuan in total. That gap makes attention a test signal before it becomes revenue.

Early access has a similar burden of proof. On Mango TV, SVIP members can watch three episodes of *The Post-Journey to the West* in advance, while VIP members can watch two. The tier difference can create urgency, but it does not establish sustained willingness to pay for AI drama.

Public response also needs to be read as mixed evidence rather than a conversion forecast. According to reports, as of September 1 the event produced 437 items of online public-opinion information: 122 positive, 92 negative, and 223 neutral. Positive information represented 27.9%, compared with 21.1% negative and 51.0% neutral. Neutrality was the largest response, while negative reaction remained substantial.

For a solo creator, the practical move is to test a short work with an immediate hook before placing a paywall or depending on platform payouts. Watch whether viewers finish it and discuss it. Repeated public interest can justify another episode or an early-access offer; a single burst of visibility cannot.

Build canonical character and scene references before asking an agent to animate anything. According to woshipm, enabling character-reference or consistency settings at 0.7-0.8 with a reference image can cut rework by 80%.

Treat each approved still as a library asset, not disposable prompt output. Seedance 2.0 is suited to 3-5-second image-to-video clips at 720P or 1080P, so assemble a story from controlled fragments. Begin with a one- to two-minute trial containing about 30-40 storyboard shots, as woshipm recommends, rather than committing to a three-minute episode that needs about 60-80 shots.

Measure retention before assuming views will fund the work. woshipm says revenue per 10,000 views fell from about 100 yuan in 2025 to 5-10 yuan by mid-2026; platform income and paywalls need evidence from the audience you actually keep.

For readers outside China

• Availability: vivago R1 is officially available globally. Its Chinese version is called Gouda (够搭). The source material does not say which countries are supported, whether access differs by region, or whether Seedance 2.0 is available outside China.
• Pricing: Pricing for vivago R1, Gouda, and Seedance 2.0 is not disclosed in sources. For adjacent production tools, one source says Midjourney costs about $10 per month, while Stable Diffusion can be used for free through local deployment; that local setup requires a dedicated graphics card with at least 8G of video memory.
• Closest Western equivalents: The source material does not identify a closest Western equivalent for vivago R1 or Gouda.; Functionally, the product combines conversational creative direction, storyboard generation, reusable asset management, and video generation; no named Western comparison is provided in the sources.
• Data residency: The source material does not cover data residency, storage location, training use of uploaded assets, retention policies, or cross-border data handling for vivago R1, Gouda, Seedance 2.0, or Mango Lingchuang.

Sources

• woshipm AI长剧的尴尬：降本容易赚钱难 https://woshipm.com/ai/6462878.html
• ifanr 实测全球上线的 vivago R1，我用它打造了一部赛博西游记短片 https://ifanr.com/1678636
• 36kr 芒果超媒20CM涨停背后：如何从企业动态中提前发现市场信号？ https://36kr.com/p/3964575113436680
• woshipm AI漫剧从0到1：零基础怎么做出一部能赚钱的漫剧？（保姆级教程） https://woshipm.com/ai/6461537.html