# Generate a short clip or b-roll shot from a prompt

> Generate a short clip or b-roll shot from a prompt: 0 companies fully do it, 14 with a limit, 34 carry the tag. Every listing in Video generation with the reason and a source.

- Canonical: https://www.anchorterminal.com/niches/video-generation.text-to-clip/
- Markdown: https://www.anchorterminal.com/niches/video-generation.text-to-clip/index.md (~1,750 tokens)
- Slim: https://www.anchorterminal.com/niches/video-generation.text-to-clip/index.min.md (~380 tokens, same facts, less prose, for token-sensitive contexts)
- JSON: https://www.anchorterminal.com/niches/video-generation.text-to-clip/index.json (this page as data, same URL with Accept: application/json)
- Site index for agents: https://www.anchorterminal.com/llms.txt (the key docs and every listing in one file, at most 16 MB, about 4 million tokens: https://www.anchorterminal.com/llms-full.txt)
- API: https://www.anchorterminal.com/api/v1/index.json
- MCP: https://www.anchorterminal.com/mcp (the whole catalogue as tools, no key)
- DNS: dig +short TXT <slug>.grade.anchorterminal.com (the grade as a DNS TXT record, for a sandbox with no HTTP where names still resolve; also top., best.<category>. and <a>.vs.<b>. under the zone, and the same under grade.letme.dev)
- Licence: CC BY 4.0, credit Anchor Terminal and link this page (https://www.anchorterminal.com/terms/#data)
- Updated: 2026-11-15

No company's docs show a full fit yet. The listings below do the job with a limit, or not at all.

| Companies, full fit | With a limit | Carry the tag | Share of the category's work | First-hand accounts |
| --- | --- | --- | --- | --- |
| 0 | 14 | 34 | 16 | 8 |

## What counts as done

An MP4 at the requested duration, ratio and resolution, downloaded to the agent's storage, from one create call and a poll or callback, with the cost known.

Measured, for our own tests: within 10 minutes.

How agents ask for it: make a five second clip of; generate b-roll for this script; make a video of a cat surfing; give me a cinematic shot of

## Does the job, with a limit

It does the job, but something in the docs or a user's account will get in an agent's way. The limit is named in each row.

- **fal video models (Features & Labels, Inc.)**: One queue submit returns request_id, status and result URLs, with fal_webhook as a callback and duration, resolution and ratio per model, but no job times are published so the 10 minute bar is unknown.
- **framefields (Gatewai)**: Prompt-to-clip runs only through the optional credit-billed AI gateway, whose per-model prices are drawn by script; the library itself renders from code.
- **Google Veo (Google LLC)**: 4, 6 or 8 second clips at 720p, 1080p or 4K in 16:9 or 9:16 from a long-running operation, but first-hand reports show no single client completing create then poll, a GA model routed to v1beta1, and a download URL returning 403 project mismatch.
- **Higgsfield API (Higgsfield Inc.)**: One submit with hf_webhook or polling status_url, duration per model, a pre-submit estimate and seven-day retention with expires_at, but no job times are published so the 10 minute bar is unknown.
- **Luma AI API (Luma AI, Inc.)**: One create with duration 5s or 10s and resolution, then poll GET /v1/generations/{id}, with credits_charged in the result and a per-generation expires_at, but there is no webhook and no job times are published.
- **Magic Hour MCP (Magic Hour AI, Inc.)**: create_video_from_text sets end_seconds, aspect_ratio and resolution, wait_for_video_project polls up to a set timeout and returns signed download links, and creation returns credits charged, but the credit price before the call isn't in the tool schema.
- **MiniMax Video API (Nanonoble Pte. Ltd.)**: One create with duration (4 to 15 s), resolution and ratio, then poll or callback_url, and usage seconds in the task; but output URLs are time-limited with no stated length and no job times are published against the 10 minute bar.
- **Pika API (Mellis, Inc.)**: Submit returns a job, then poll /v1/media/jobs/{id} or take a signed webhook, with duration and resolution in the schema and a free quote, but no job times are published so the 10 minute bar is unknown.
- **PixVerse API (MOTIVAI PRIVATE LIMITED)**: Create with duration 1 to 15 s on C1 and V6 and quality, then poll the result or use a configured webhook_id, but credit prices are pegged in docs while plan prices sit on a JavaScript-only page, and no job times are published.
- **Popcraft MCP (Popcraft)**: Jobs return an id at once with get_cost: true pricing before submit and a self-refreshing result card; the README says video takes a few minutes, but durations, retention and download contract weren't verifiable in the tool schemas.
- **Replicate video models (Replicate, LLC)**: POST a prediction with duration and aspect ratio, then poll or set webhook, but outputs and inputs are deleted after an hour by default, no job times are published, and per-model prices sit on JavaScript-drawn model pages.
- **Runway API (Runway AI, Inc.)**: One create, then poll GET /v1/tasks/{id} or the SDK wait helper, with ratio and duration per model and a published price table, but there is no webhook, job time is stated only as seconds to minutes, and output URLs expire in 24 to 48 hours.
- **Seedance on BytePlus ModelArk (BytePlus Pte. Ltd.)**: One create task with duration or frames, ratio and resolution up to 4K, then poll or callback_url, with a token-based price formula, but no job times are published and a task may run until its 48 hour default expiry.
- **Vidu API (ShengShu AI HK Limited)**: One create with duration 1 to 16 seconds and 540p to 1080p, then poll task detail or callback_url, with credits in the response, but no job times are published and off-peak jobs may wait up to 48 hours.

## Couldn't check

The docs were closed to us, or silent. These carry no mark either way.

- **Alibaba Wan (Model Studio) (Alibaba Cloud (Singapore) Private Limited)**: alibabacloud.com did not resolve from here, so the Wan 3.0 API's async contract, durations and retention couldn't be read.
- **Kling AI API (Kling AI Pte. Ltd.)**: Create and poll endpoints, durations and callback options sit in the JavaScript-only reference; only per-second prices were readable.

- not this job: Clueso (Desklamp, Inc.): No prompt-to-video model; projects are built from slides, recordings, elements and AI images, then exported.
- not this job: HyperFrames (HeyGen): No model-generated clip from a prompt; the agent writes HTML that the CLI renders.
- not this job: sync. labs MCP server (Synchronicity Labs): No prompt-to-video generation; tools lip-sync, dub and edit dialogue in existing footage.

1 listings in the category were added after this reading and aren't marked yet.

## What stops agents finishing

- create and poll calls need different credentials so no single client finishes the job
- result URL needs a key from a project the caller does not own
- model id retired or moved to another endpoint version without notice
- no API access yet for the model the user named

## Around it

Before: llm.write-script. After: video-generation.code-rendered-video, storage.upload, social.publish.

## Where this comes from

Weight 16 of 100 in Video generation: 8 of 37 accounts; b-roll for explainers, scripts calling Kling and Runway directly, and the Veo SDK and download failures

As data: https://www.anchorterminal.com/api/v1/use-cases/video-generation.text-to-clip.json, or the MCP tool get_use_case with id video-generation.text-to-clip. Draft, read 2026-11-15 from public docs and first-hand accounts, not yet reviewed by a person.

---

Query this catalogue directly over MCP: https://www.anchorterminal.com/mcp (Streamable HTTP, no key). How: https://www.anchorterminal.com/agents/#mcp
