Seedance multi-version generator (default 2.0)

Generate video with Seedance 2.0, Seedance 1.5 Pro, or Seedance 1.0 variants on WeryAI. model is required in every JSON body—default model_key is SEEDANCE_2_0 unless the user explicitly picks another row in ## Model and API constraints.

Dependencies: WERYAI_API_KEY + Node.js 18+. When you run the CLI, {baseDir}/scripts/video_gen.js must exist; WERYAI_VIDEO_API.md must exist under {baseDir}/resources/ (publish or pre-use assembly). Full commands and JSON fields: see resources/WERYAI_VIDEO_API.md. No other Cursor skills required.

Choosing a Seedance version (quick map)

Always re-check duration, aspect_ratio, resolution, generate_audio, and whether negative_prompt is allowed for the selected model_key and channel (text vs image)—see frozen tables below.

Prerequisites

• WERYAI_API_KEY must be set before running video_gen.js.
• Node.js 18+; reference URLs must be public https (no local paths).
• model is mandatory in wait --json. If the user does not specify a version, use SEEDANCE_2_0.
• Each wait run may consume credits; re-run creates new paid tasks.

Security, secrets, and API hosts

• WERYAI_API_KEY: Treat as a secret. Only configure it if you trust this skill's source; it is listed in OpenClaw metadata as requires.env / primaryEnv so installers know it is mandatory at runtime (never commit it inside the skill package).
• API hosts (fixed in video_gen.js): Video tasks use https://api.weryai.com; the models list uses https://api-growth-agent.weryai.com. Child skills must not document or rely on URL environment-variable overrides—only WERYAI_API_KEY is read from the environment.
• Higher assurance: Run generation in a short-lived or isolated environment (separate account or container), and review scripts/video_gen.js (HTTPS submit + poll loop) before production use.

Prompt expansion (mandatory)

video_gen.js does not expand prompts. Before every wait --json, turn the user's short brief into a full English production prompt.

When: Input is short or vague. Exception: The user supplies a finished long prompt within prompt_length_limit and explicitly asks you not to rewrite—still show the full text in the confirmation table.

Always add: shot size and angle; camera move or lock-off; lighting and mood; motion paced to duration; one clear payoff; framing for aspect_ratio; if generate_audio is true, describe ambience/SFX in generic words (no copyrighted music).

negative_prompt: Send only when the frozen row for that model_key and channel shows Yes. Otherwise omit the field.

Length: Stay within the selected row's prompt_length_limit (all listed: 2000).

Confirmation: The pre-submit table must include the full expanded prompt, model, and every field you will send.

Workflow

1. Detect whether the job is text-to-video or image-to-video (including multi-image / first-last).
2. Pick model_key: default SEEDANCE_2_0; if the user names another Seedance line, map with ## Choosing a Seedance version and the frozen tables.
3. Validate features vs model: e.g. do not send images with more than one URL for models with support_multiple_images: false or upload_image_limit: 1.
4. Expand prompt (mandatory) per above.
5. Show the confirmation table (full prompt, model, duration, aspect_ratio, resolution, generate_audio, optional negative_prompt, URLs).
6. Run node {baseDir}/scripts/video_gen.js wait --json '...'.
7. Return videos as Video Markdown links—not inside code fences.

CLI reference

~~~sh

node {baseDir}/scripts/video_gen.js models --mode text_to_video

node {baseDir}/scripts/video_gen.js models --mode image_to_video

node {baseDir}/scripts/video_gen.js wait --json '{"model":"SEEDANCE_2_0","prompt":"…","duration":10,"aspect_ratio":"9:16","resolution":"720p","generate_audio":true}'

~~~

Full reference: resources/WERYAI_VIDEO_API.md.

Definition of done

Done when the user gets a playable Video link or a clear error with a parameter fix. The submitted payload must match the selected model_key row. Default version behavior: SEEDANCE_2_0 when unspecified.

Boundaries (out of scope)

• Do not send negative_prompt when the frozen row says No for that model and channel.
• Do not exceed upload_image_limit or use multi-image when support_multiple_images is false.
• Do not use local file paths for media; do not embed WERYAI_API_KEY values in files.
• Do not link to weryai-model-capabilities.md or arbitrary ../references/ paths; use resources/WERYAI_VIDEO_API.md for CLI/API details.
• Do not wrap user-facing playable URLs in Markdown code fences.

Example prompts

• Default Seedance vertical: rainy alley, slow push-in, 10s → SEEDANCE_2_0
• Use Seedance 1.5 Pro, 1080p, 3:4, text-only → DOUBAO_1_5_PRO
• Seedance 1.0 Lite, three product photos in order → DOUBAO_1_LITE + images (≤3 URLs)
• First and last frame transition, best quality → prefer SEEDANCE_2_0; DOUBAO_1_5_PRO only if user insists on 1.5 and accepts single reference constraints (see table)

Model and API constraints (frozen for this skill)

> From node {baseDir}/scripts/video_gen.js models (text_to_video + image_to_video) at authoring time. Re-run after upgrades. multi_image_to_video mode returned [] on the authoring run—use image_to_video rows below for image counts and flags.

Text-to-video

Image-to-video

Recommended defaults (when user is silent)

Scenario notes

• Text-only: Any row in the text table; default SEEDANCE_2_0.
• Single image: Any image row; respect upload_image_limit.
• Several images / storyboard: Prefer SEEDANCE_2_0 or DOUBAO_1_LITE—never send multi-image payloads to models with Multi-image = No.
• First + last frame: SEEDANCE_2_0 or DOUBAO_1_5_PRO per image metadata; confirm against WERYAI_VIDEO_API.md field names (first_frame / last_frame, image / last_image, or ordered images).

> Pack: seedance-video-gen-1-2.0.