Capabilities · Last updated September 9, 2026
Video generation for AI agents.
Give an agent a prompt or source image, choose a model, and get a clip back without writing a separate integration for every provider. The CLI also exposes editing models, motion control, live modes, and credit estimates, so the agent can make a valid choice before it submits the job.
When is this useful?
Use it when video is one step in a larger job—for example, turning a product brief into a clip, reviewing the result, and sending the approved file to a page or Drive. Query the live catalog first because models do not all support the same inputs.
- Text-to-video, image-to-video, and edit-video use one command surface.
- Live catalog modes and credit estimates stay visible before an agent chooses.
- Featured decisions explain both best fit and the practical trade-off.
- Returned clips stay available for review, delivery, and the next agent step.
Checked against the live catalog · Catalog verified September 9, 2026
Choose video models by their supported inputs.
Our September 9, 2026 catalog check with AnyCap CLI 0.6.1 returned 16 active video models, compared with 13 in the August 5 snapshot. This verifies catalog availability and declared modes, not generation quality. Use the current listing below for model selection and query the live schema before each run.
- September 9 snapshot
- 16 active models
- Generation
- text, image, first/last frame, multi-shot
- Editing
- edit-video and motion-control where declared
- CLI version
- 0.6.1
anycap video modelsModels grouped under video generation
This is the current category view, not a permanent shortlist. Use it to compare model families and input modes, then check the live schema before sending a production request.
- Seedance 2.5text-to-video, image-to-video, first-last-frame-to-video, multi-modal-reference
- MiniMax H3 Maxtext-to-video, image-to-video, first-last-frame-to-video
- MiniMax H3text-to-video, image-to-video, multi-modal-reference
- Kling 3.0 Motion Controlmotion-control
- Seedance 2.0 Fasttext-to-video, image-to-video
- Seedance 2.0 Minitext-to-video, image-to-video, first-last-frame-to-video, multi-modal-reference
- Kling 3.0text-to-video, image-to-video, first-last-frame-to-video, multi-shot-video
- Seedance 2.0text-to-video, image-to-video, first-last-frame-to-video, multi-modal-reference
- Gemini Omni Flash Previewedit-video
- Kling 3.0 Omnitext-to-video, image-to-video, multi-shot-video
- Hailuo 2.3text-to-video, image-to-video
- Kling O1image-to-video
- Seedance 1.5 Protext-to-video, image-to-video
- Sora 2 Protext-to-video, image-to-video
- Veo 3.1text-to-video, image-to-video
- Veo 3.1 Fasttext-to-video, image-to-video
Models change. AnyCap keeps discovery, supported modes, authentication, execution, and output delivery behind one agent-facing interface, so an agent can pick a valid option today and switch when the catalog changes.
CLI usage
Run anycap video models for the full live inventory, then inspect the selected model schema before supplying local references. These examples show the current command shape for each input path.
Text-to-video
anycap video generate --prompt "a drone shot flying over a mountain range at sunset" --model veo-3.1 -o hero.mp4Image-to-video
anycap video generate --prompt "animate this still into a subtle camera move" --model seedance-2.5 --mode image-to-video --param images=./frame.jpg -o animated.mp4Edit existing video
anycap video generate --prompt "brighten the lighting and add a subtle slow zoom" --model gemini-omni-flash-preview --mode edit-video --param videos=./source.mp4 -o refined.mp4Discover models
anycap video modelsStart from the input you actually have
text-to-video
Text brief
Start from a written scene, subject, motion, camera, and output brief.
Not appropriate when an exact source frame must be preserved.
image-to-video
Reference image
Animate a prepared still while using the prompt to direct motion and camera behavior.
Requires a supported model and a prepared source still.
edit-video
Existing footage
Refine a source clip with instructions instead of generating a new clip from scratch.
Available only when the selected live model exposes that mode.
multi-modal-reference
Multimodal references
Use a guiding prompt with image, video, or audio references when one source frame is not enough to describe the intended clip.
Inspect the selected model schema because supported reference fields and limits vary by model.
Choose the right video model at runtime
Seedance 2.5 →
Use Seedance 2.5 on AnyCap for text-to-video, image-to-video, first-and-last-frame control, or multimodal reference generation through one authenticated agent workflow.
Featured on AnyCap · Active model ID: seedance-2.5
MiniMax H3 →
Use text, image, video, or audio references through the active AnyCap multimodal video model.
Also live · Runtime ID and modes verified
Five featured model decisions — Seedance 2.5 first
| Model | Modes | Cost | Best fit | Trade-off |
|---|---|---|---|---|
| Seedance 2.5 | text-to-video, image-to-video, first-last-frame-to-video, multi-modal-reference | Varies by catalog pricing | Use it as the featured AnyCap video model for text, image, first/last-frame, and multimodal reference workflows. | Inspect the live schema before sending references because accepted fields vary by mode. |
| MiniMax H3 | text-to-video, image-to-video, multi-modal-reference | Varies by catalog pricing | Use it when the agent needs one model for text, image, video, or audio-referenced generation through the AnyCap CLI. | Inspect the live schema before supplying references because fields and limits vary by mode. |
| Gemini Omni Flash Preview | edit-video | Varies by catalog pricing | Use it when the agent starts with an existing clip and needs natural-language video edits, refinements, or review loops without leaving the AnyCap workflow. | It edits existing footage only; use a generation model for a new clip. |
| Kling 3.0 | text-to-video, image-to-video | ~9 credits / second | Use it when the brief depends on camera movement, character motion, or exploratory image-to-video iteration. | Choose a lower-cost or faster model when preview speed matters more than cinematic motion. |
| Veo 3.1 | text-to-video, image-to-video | ~20 credits / second | Use it when the first pass needs to feel more polished from a text brief, especially for branded teasers and concept clips. | Choose Veo 3.1 Fast when iteration speed matters more than maximum polish. |
Generation time varies by model, queue state, duration, and settings. The CLI waits for server-side completion and returns the result when the command finishes; no fixed latency is promised.
When agents need video generation
A reliable run starts with the catalog, not a remembered model name: check the available modes, submit a supported request, review the returned clip, and only then pass the file to the next step.
Launch and demo clips
Turn product briefs, changelog notes, or storyboard prompts into short launch clips from the agent session.
Reference frame to motion
Animate design stills, screenshots, generated images, or product frames without changing providers or auth flows.
Agent-led social content
Generate short-form clips for campaigns, creator workflows, and product announcements, then hand off the asset through Drive or Page.
Video workflow prototyping
Test model fit, prompt structure, duration, aspect ratio, and reference-image behavior before building a deeper integration.
Workflow fit signals
- Best fit
- Agents that need repeatable video creation or editing inside coding, research, or content workflows.
- Not a fit
- Teams that only need a consumer editor with a manual, timeline-first interface.
- Operational check
- The agent can inspect the live model schema, retain request context, and handle delivery separately from generation.
What common failures mean
The selected model does not expose the requested mode
Run anycap video models <model-id> schema --operation generate --mode <mode> and choose a mode the live model declares.
A local reference is missing or uses the wrong schema key
Inspect the live schema and use --param images=... or --param videos=... only when that parameter is declared.
The automatic download fails after generation
Generation may already have succeeded, but the CLI reports the automatic-download error before the normal result. Do not assume a URL or request ID was returned. Fix the local path or connection, retain trace_id for support, and rerun deliberately because another clip may be created.
The output does not match the brief
Review the clip, tighten motion, camera, and subject instructions, then retry without claiming deterministic frames.
FAQ
What does AnyCap video generation let agents do?
It gives agents one command surface for text-to-video, image-to-video, model discovery, generated asset handling, and downstream sharing. Teams can generate new clips, animate reference images, and keep video creation inside the same agent workflow.
Which video models are available through AnyCap today?
Run anycap video models for the full live inventory. This page highlights four useful decisions, but availability, modes, schemas, and catalog pricing remain authoritative in the CLI.
Why highlight Gemini Omni Flash Preview on a video generation page?
The current AnyCap catalog exposes Gemini Omni Flash Preview as an edit-video model. Use it after an agent has a source clip to refine, and choose a text-to-video or image-to-video model when the workflow needs a new clip.
Why does this page include image-to-video as well as text-to-video?
Agent workflows often start from a screenshot, design frame, or product still rather than a prompt alone. AnyCap treats both text-to-video and image-to-video as one video generation capability so the workflow stays consistent.
Is this page about a video generation API or a CLI workflow?
It covers both the video generation capability and the CLI workflow agents use to operate it. AnyCap provides model discovery, live schemas, generation commands, and result delivery through one CLI-first surface over multiple video models.
Why use AnyCap instead of wiring one video provider directly?
A direct provider integration can work when the agent always uses one model. AnyCap is better when the agent needs model choice, shared auth, CLI execution, video review, file handling, and a path to publish or share outputs without rebuilding every workflow per provider.
Let your agent generate the clip.
Start with the CLI, choose a live model and supported mode at runtime, then keep generation, review, and delivery in the same agent workflow.