UseTrawsnUseTrawsn
Capabilities

Generate

Image GenerationCreate and edit images from prompts or references.Video GenerationCreate motion outputs from text and image inputs.Music GenerationProduce music tracks through one runtime.Audio GenerationGenerate speech, dialogue, sound effects, and complete audio scenes from text, audio, or image input.

Understand

Image UnderstandingRead screenshots, diagrams, and visual references.Video AnalysisInspect recordings and extract structured details.Audio UnderstandingTranscribe and analyze voice and audio files.

Retrieve

Web SearchSearch the web from the same agent workflow.Grounded Web SearchReturn synthesized answers with live citations.Web CrawlFetch pages and convert them into clean content.Social Media LookupRead Instagram and X profiles, posts, feeds, and X search results.

Store

DriveStore outputs, organize assets, and create public URLs.
Equip Agents
Claude CodeCursorCodexDeepSeek HarnessManus
Resources

Explore

GuidesDecision guides for building reliable agent workflows.Context EngineeringUnderstand how prompts, files, and workspace state shape agent behavior.Agent SkillsSee how reusable skills package workflows and capability usage for agents.

Evaluate

Compare UseTrawsnBrowse comparison pages for adjacent agent tooling, media APIs, and tradeoffs.GlossaryA shared vocabulary for agent capabilities, tools, and workflows.
Docs ↗Pricing
I'm Agent
I'm Agent
  1. Home
  2. Equip Agents
  3. Claude Code
  4. Video Generation

Claude Code x Video Generation · By UseTrawsn Team · Updated November 16, 2026

Can Claude generate videos? Not on its own. Here's how to do it from Claude Code.

Claude can't render video. What Claude Code can do is plan the clip, write the prompt, prepare reference images, and decide what happens to the result. To get the actual file, it calls a video model such as Seedance 2.5, Veo 3.1, or Kling 3.0 through UseTrawsn. Claude Code orchestrates the work, the selected model renders it, and the live UseTrawsn schema defines which inputs are valid.

Get StartedAgent skills guide

What Claude can and can't do with video

Anthropic's model docs list text as the only output, and its upload docs don't include any video format. So it depends on what kind of video you need.

What you wantClaude on its ownClaude Code + UseTrawsn
Footage from a text promptNo. Anthropic has no video model.Text-to-video with an active model such as Seedance 2.5, Veo 3.1 or Kling 3.0.
Animate a still image or follow a reference clipNo.Image-to-video, first-last-frame, multimodal-reference or motion-control modes, where the model's schema supports them.
Motion graphics, captions, slideshows or UI animationYes, as code: HTML and CSS or SVG animation, a Remotion project, or an ffmpeg script that Claude Code writes and runs.Same code route. Add generated clips when a scene needs real footage.
Check a finished clipNo. Claude can't watch a video file.usetrawsn actions video-read returns a text analysis that Claude Code can act on.

Checked November 2, 2026 against Anthropic's models overview and upload documentation.

One real clip, from request to checked file

One Kling 3.0 text-to-video job, followed through the request, the queued task, the resumed download, the file, the actual charge and the checks. Every value below comes from that single production run on September 29, 2026. Only the task ID is replaced with a placeholder.

Who ran it: the Codex desktop host, using UseTrawsn CLI 0.6.3 against the production API (api.usetrawsn.ai). This is not a Claude Code recording, and we have not recorded a Claude Code session for this page. The commands are the same ones Claude Code runs through the UseTrawsn skill; only the host that issued them differs.

The MP4 the resumed task downloaded, published unedited (byte-identical). Kling 3.0 text-to-video, 5 s, 1080p, 16:9, generate_audio=false, so it is silent by request.

Submit the request (endpoint flag removed)

$
usetrawsn video generate --model kling-3.0 --mode text-to-video --prompt 'A single continuous 5-second photorealistic product shot. A matte cream product box stands upright on a clean oak desk, viewed from a slight three-quarter angle. Soft daylight from a window on the left, gentle realistic shadows, neutral softly blurred office background. The camera slowly and smoothly pushes straight in toward the box throughout the shot, moving from a medium shot to a closer product view. The box and desk remain stationary with consistent geometry. No cuts, no camera shake, no people, no added text or logos.' --param duration=5 --param aspe<task_id>=16:9 --param resolution=1080p --param format=mp4 --param generate_audio=false -o codex-demo.mp4 --no-wait

--no-wait returns as soon as the task is accepted instead of blocking until the file is ready. Every parameter used here is still in the production Kling 3.0 text-to-video schema we rechecked on November 16, 2026 (duration 3–15 s; 16:9, 9:16 or 1:1; 720p, 1080p or 4k).

Response after 1.0 s · excerpt, task ID replaced

{
  "data": {
    "status": "queued",
    "poll_after_ms": 2000,
    "interaction_mode": "async",
    "client_surface": "cli",
    "task_id": "<task_id>"
  }
}

Queued means accepted, not finished: no video exists yet. The task ID is the checkpoint. Save it before the session ends so the job can be resumed instead of submitted and billed again.

Resume and download · 59.5 s, status success

$
usetrawsn video tasks wait <task_id> -o codex-demo.mp4

This waited for the render, then saved the output as an MP4. A later Claude Code session can run the same command with the saved ID; usetrawsn video tasks get <task_id> reports the status without waiting.

What it cost and how it was checked

Charge
17 credits for the generation, reported by the CLI when the task succeeded. That is $0.34 at the Starter pack rate ($10 for 500 credits); larger credit packs cost less per credit. One observation for Kling 3.0 at 5 s and 1080p without audio, not a price quote.
With review
The video-read review below is a fixed 10 credits per request, so generating and reviewing this clip came to 27 credits ($0.54 at the Starter rate). No paid retries were needed.
Other models
Prices differ by model and settings: a separate Seedance 2.5 trial (5 s, 720p, no audio) reported 59 credits.

File specs · ffprobe, re-run November 16, 2026

ffprobe -v error -show_entries stream=codec_type,codec_name,width,height,r_frame_rate:format=duration,size -of compact codex-demo.mp4

stream|codec_name=h264|codec_type=video|width=1920|height=1080|r_frame_rate=24/1
format|duration=5.041667|size=5402637

There is one stream, and it is video, so the file has no audio track, as requested. We ran this check on the published copy, which is byte-identical to the downloaded file.

Optional review · usetrawsn actions video-read, about 18 s, 10 credits

$
usetrawsn actions video-read --file ./codex-demo.mp4 --instruction 'Respond entirely in English. Analyze this video for three things: (1) Describe the camera movement throughout, distinguishing camera motion from product motion. (2) Is any text on the product box legible? Transcribe only text you can confidently read and explicitly identify uncertain or unreadable text without guessing. (3) Describe visible artifacts, including warping, flicker, jitter, inconsistent geometry, text distortion, blur, or other temporal inconsistencies. Give approximate timestamps where useful, distinguish observations from uncertainty, and keep the response concise.'

The analysis model described a smooth, continuous push-in with the box stationary and found no text on the box. It flagged subtle warping along the top flap folds at about 00:02–00:05 and saw no pronounced flicker or jitter. Its metadata header (1920 × 1080, 24 fps, 5.0 s, no audio) matches ffprobe. Read the full, unedited response.

Frame at 0 seconds: a blank cream box on an oak desk, with a window and two chairs in view
Frame at 0 s
Frame at 4 seconds: the same blank box, framed noticeably closer, with the chairs nearly out of view
Frame at 4 s

Our own frame check: the box is blank in both frames and larger at 4 s, and the chairs have nearly left the frame. Two stills can't confirm or rule out the subtle seam warping the analysis model reported, so watch the clip around 00:02–00:05 to judge it yourself.

Kling 3.0 model pageVideo analysis capability

Choose the workflow from the inputs you have

InputsWorkflowReview before delivery
A written scene briefText-to-videoSubject, motion, duration and unwanted text.
An approved still imageImage-to-videoSubject continuity, composition and camera movement.
A character image and a motion-reference clipKling Motion ControlCharacter identity, hands, feet, occlusion and background.

For a concrete motion-transfer example, our October 21 production CLI trial used the same robot image and six-second dance clip with two orientation settings. Each output reported 23 credits. Both followed the dance, but also inherited the portrait framing of the character image and black background. This is CLI evidence, not a recorded Claude Code host session.

Watch the paired motion-control outputsPrepare or edit a still image

Install path

Step 1

$
npx -y skills add usetrawsn-ai/usetrawsn -a claude-code -y

Step 2

$
curl -fsSL https://usetrawsn.ai/install.sh | sh

Step 3

$
usetrawsn login

What Claude Code can verify before it spends a generation request

The model name is not the contract. A reliable agent checks the runtime, discovers the current catalog, reads the selected mode schema, and only then submits the paid generation job. That sequence makes the workflow reproducible when models or accepted fields change.

  1. 1.

    Verify the runtime

    Confirm authentication and connectivity before Claude Code starts a generation task.

    $
    usetrawsn status
  2. 2.

    Discover the live catalog

    Choose from model IDs and modes returned by the current UseTrawsn catalog rather than relying on a remembered model list.

    $
    usetrawsn video models
  3. 3.

    Inspect accepted inputs

    The schema is the contract for mode-specific fields. Claude Code should inspect it before composing parameters or references.

    $
    usetrawsn video models seedance-2.5 schema --operation generate --mode text-to-video
  4. 4.

    Generate an artifact

    A successful task returns a real video artifact that Claude Code can inspect, publish, or pass to the next workflow step.

    $
    usetrawsn video generate --model seedance-2.5 --mode text-to-video --prompt "controlled product reveal, slow dolly in, soft studio lighting" -o launch-teaser.mp4
  5. 5.

    Resume a pending task or retry delivery

    Keep the task_id emitted when generation is accepted. If the wait is interrupted, times out locally, or the download fails, resume this existing task to wait and download its result. Use usetrawsn video tasks get <task-id> for one status check; do not submit the same paid generation again.

    $
    usetrawsn video tasks wait <task-id> -o launch-teaser.mp4

What this workflow does not guarantee

  • Claude Code plans and invokes the workflow; the selected video model performs the rendering.
  • Supported references, duration, resolution, and output fields vary by model and mode. Inspect the live schema before each new workflow shape.
  • A text-to-video prompt does not guarantee exact typography, identity continuity, or deterministic motion. Use references and iteration when those constraints matter.
  • Long-running generation can outlive an interactive turn. Preserve the task or output path so the workflow can recover without submitting the same paid job twice.

First command to try

The point is not only the video call itself. It is that the agent can reach video generation through the same runtime it already uses for image and vision tasks.

$
usetrawsn video generate --model seedance-2.5 --mode text-to-video --prompt "slow camera move over a new product launch scene with soft studio lighting" -o launch-teaser.mp4

Finish with a playable file

  • Keep the returned task ID and requested output filename. If the terminal stops waiting, resume that task rather than submitting another generation.
  • Open the saved MP4 and check its duration, framing, motion and audio against the brief. A completed task does not establish that the clip is suitable for publication.
  • Claude can't watch the MP4 itself. To let it review the clip, have Claude Code run usetrawsn actions video-read --file ./launch-teaser.mp4 --instruction "Does this clip match the brief? List visible defects with timestamps." and spot-check the timestamps it reports.
  • Keep the original output when trimming or editing. Confirm the final export plays and meets your destination's format and size requirements before adding it to the project.
  • Record the actual credit charge with the chosen result. The motion-control trial's 23 credits is a dated observation, not the price of every video model or duration.

When this capability matters

  • Product walkthroughs and demo clips generated directly from Claude Code workflows.
  • Launch or social visuals that need motion rather than a single frame.
  • Concept validation when the agent needs to show a flow, not just describe it.
  • A consistent expansion path from image generation into video generation without another SDK.

FAQ

Can Claude make videos?

+

Not on its own. Anthropic's models take text and images in and return text, so there's no built-in video model. Claude Code can write code that animates things, and with UseTrawsn it can call Seedance 2.5, Veo 3.1, Kling 3.0, or another video model for real footage.

Can Claude watch the video it just generated?

+

Not directly, because Claude doesn't accept video input. Claude Code can run UseTrawsn video analysis on the output file and read the returned text to decide whether the clip matches the brief.

Which video model is available today?

+

The current UseTrawsn catalog includes Seedance 2.5, Veo 3.1, Kling 3.0, and other video models. Run usetrawsn video models because availability and supported modes can change.

Why route video generation through UseTrawsn instead of a separate SDK?

+

Because the same runtime and auth flow can also cover image generation, image understanding, and video analysis without changing the way the agent works.

Back to Claude Code hubVideo Generation capabilityClaude Code image generation

Generate one video in Claude Code

A delivered video file. A pending task ID is a checkpoint, not a finished clip.

Install the CLI and sign in once. If you are already configured, continue with the access and cost checks.

$
curl -fsSL https://usetrawsn.ai/install.sh | sh
usetrawsn login
usetrawsn --version
usetrawsn status

Add the skill, then reopen your agent session if needed.

$
npx -y skills add usetrawsn-ai/usetrawsn -a claude-code -y

Run usetrawsn status, inspect the current model schema, and check pricing and your balance. Generation is billed; model access and rates can change.

Paste this prompt into your configured agent. Inspect the delivered file and actual charge.

Use UseTrawsn to create a five-second product shot of an olive-green desk lamp on an off-white desk with a slow camera push-in, no text. Check my login and the current Seedance 2.5 text-to-video schema. Use 720p, 16:9 and generate_audio=false only if the current schema supports them. Explain pricing and ask for approval before generating. Save the completed clip as ./desk-lamp.mp4, keep any task ID and resume it instead of submitting a duplicate. Report the delivered file and actual charge.

With CLI 0.6.5, expired video links are refreshed once by querying the same task, and transient downloads are retried. A pending task ID means the job is still running. Keep that ID and resume waiting after a local timeout or download failure; a new generation command creates a separate billed task.

$
usetrawsn video tasks get <task-id>
usetrawsn video tasks wait <task-id> -o desk-lamp.mp4

Read the related workflow. For setup help, see installation and login.

CLI version and updating · 0.6.5

Current release checked October 21, 2026: 0.6.5. Use the update command for your installation method, then verify the version and login in the same environment that runs your agent.

Direct binary installation

$
usetrawsn update

npm installation

$
npm install -g @usetrawsn/cli@latest
$
usetrawsn --version
usetrawsn status

Version 0.6.5 retries transient downloads, recovers an expired video link by querying the same task, and saves multiple image outputs to numbered files. After a video download failure, resume the existing task instead of generating again.

$
usetrawsn video tasks wait <task-id> -o result.mp4

Dated examples retain the CLI version, output and charge from the original run. Updating the CLI does not revalidate an old result. Check the live model schema and current pricing before a new billed task.

Capabilities

  • Overview
  • Image Generation
  • Video Generation
  • Music Generation
  • Image Understanding
  • Video Analysis
  • Audio Understanding
  • Web Search
  • Grounded Web Search
  • Web Crawl
  • Social Media Lookup
  • Drive

Equip Agents

  • Overview
  • Start here
  • Claude Code
  • Cursor
  • Codex
  • Manus

Resources

  • Overview
  • Context Engineering
  • Agent Skills
  • What Agents Can't Do
  • Compare agents and tools

Product

  • Product overview
  • Models
  • Install UseTrawsn
  • Install the Agent Skill

Documentation

  • Docs overview
  • Install UseTrawsn
  • MCP setup
  • CLI reference

Company

  • About
  • Contact
  • Privacy
  • Terms
UseTrawsn
Join the UseTrawsn Discord