dreamina-canvas-generate-video

Dreamina Canvas Video Node

This Skill owns the Canvas video node. It exists so that video references, modes, and ratio constraints are correct without confusing i2v (which does not exist) with m2v, or multi_modal (a server internal alias) with the public mode set.

Reference detail is in references/video-node-contract.md.

When to use

• The caller wants to save a video draft (t2v, first_last_frame, or m2v).
• The caller wants to edit an existing video node, including replacing the entire generation block.
• The caller wants to switch modes on an existing draft and needs the reference rules explained.

When not to use

• For approving or running generation — that is dreamina-canvas-quote-and-run.
• For the audio track of a video — that is dreamina-canvas-generate-audio.

Modes — the public set

t2v               text-to-video
first_last_frame  one or two ordered frames
m2v               image / video / Element / multimodal guidance

There is no i2v mode. The server-internal alias multi_modal must not be passed; the CLI rejects it with exit code 2. Image-to-video, video-guided generation, Element references, and "single image to video with a preserved ratio" all use m2v.

# t2v
dreamina-canvas --format json node create video \
  --prompt "<prompt>" \
  --mode t2v --model <m> --ratio <r> --resolution <res> --duration 5

# first_last_frame (one frame)
dreamina-canvas --format json node create video \
  --prompt "<transition prompt>" \
  --mode first_last_frame --model <m> \
  --ref node:<firstFrameNodeId> --duration 5

# first_last_frame (two frames)
dreamina-canvas --format json node create video \
  --prompt "<transition prompt>" \
  --mode first_last_frame --model <m> \
  --ref node:<firstFrameNodeId> --ref node:<lastFrameNodeId> --duration 5

# m2v (image / video / Element guidance)
dreamina-canvas --format json node create video \
  --prompt "<motion prompt>" \
  --mode m2v --model <m> \
  --ref node:<imageOrVideoNodeId> --duration 5

--duration

Mandatory for any video draft in seconds, must be > 0, default 5. Take the upper bound from the discovered model spec; do not invent.

--ratio

Some video models reject an explicit --ratio (for example, image- and frame-driven modes follow the source frame ratio). Discovery wins; if the model spec says the ratio is inferred from the first frame, do not pass --ratio.

first_last_frame rules

• One or two --ref entries are allowed.
• When two frames are supplied the source ratios of the two images must match.
• The output ratio is determined by the first frame; passing two frames with mismatched ratios is rejected by the server.

Reference rules

Same surface as dreamina-canvas-generate-image:

• node:node_xxx — follow the node (canvas edge).
• res:<lowercase UUID> — freeze to a specific resource.
• uri: / vid: — reserved, rejected.
• Internal short links (video_xxx, image_xxx) — rejected.

Video nodes may reference Image, Element, or Video nodes/resources.

Generation is a full replace

Identical to image:

• Touching --mode, --model, --ratio, --resolution, --duration, --count, --prompt, --ref requires the complete new block.
• Metadata is sparse; only the flags you pass are updated.
• --clear-generation is the opt-out (mutually exclusive with generation flags).

Paid execution is not this Skill's job

Hand off to dreamina-canvas-quote-and-run. This Skill never calls --run, never sets --credit-ceiling or --credit-token, never persists tokens.

Failure → recovery

Failure	requiredAction	What to do next
--mode i2v rejected (exit 2)	none	Switch to m2v.
--mode multi_modal rejected (exit 2)	none	Switch to m2v.
Two-frame first_last_frame ratios disagree (exit 2)	none	Pick two frames with matching source ratios.
Explicit --ratio rejected by the model (exit 2)	none	Re-run discovery; remove --ratio if the model infers from the first frame.
Missing --duration (exit 2)	none	Pass --duration (default 5, model bound applies).

What this Skill will not do

• Approve spend.
• Persist tokens, signed URLs, cookies, or session material.
• Accept i2v or multi_modal as a mode.
• Treat generation edits as a partial update.

Policy

Invocation requires:

• Confirmed installation of dreamina-canvas
• Live model discovery result for video
• For paid execution: explicit handoff to dreamina-canvas-quote-and-run

Forbids:

• Passing --mode i2v or --mode multi_modal
• Calling --run from this Skill
• Supplying --credit-ceiling or --credit-token
• Persisting tokens, signed URLs, cookies, or session material

Default prompt:

Public video modes are t2v, first_last_frame, m2v. There is no i2v and no multi_modal --mode value. When the user says 'image to video' or wants to preserve a ratio, use m2v. Hand off paid execution.