Phenomenal AI

Showing 25 of 27 from the live probe.

• Create Storyboard

  Create an empty storyboard project for planning a multi-shot sequence. This does not generate images or video. Do not create a storyboard for a simple single-image request.

• Get Storyboard

  Read an owned storyboard and its current shots, sheets, and generation state. Use to inspect a plan before changing or generating it.

• Generate Shots From Script

  Append a shot plan from a script to an existing storyboard. Set autoGenerateImages=false for planning only. Otherwise image generation starts automatically and can charge credits; autoGenerateVideos also starts video generation when configured. The returned Director task ID is not a completed film or an ordinary media task; inspect get_storyboard for persisted results.

• Generate Shot Image

  Generate or regenerate one existing storyboard shot image. May charge credits. Select an available text-to-image or image-edit model and its supported settings. Returns a generation task ID; use check_task_status for completion.

• Approve Shot

  Approve an existing shot that is pending approval. This changes storyboard state; it does not generate or select a completed image. The legacy generationId argument does not select an output.

• Get Current Plan

  Get the authenticated user plan, credits and generationLimits before paid work. Use generationLimits.video for account duration policy; plan.features.maxVideoLength is legacy metadata, not an enforced universal clip limit. Model and workflow constraints still apply.

• Get Plan Restrictions

  Get the authenticated user quality restrictions and generationLimits.video account duration policy. Do not infer a universal ten-second limit from legacy plan metadata. Model availability does not by itself guarantee entitlement or sufficient credits.

• Get My Gallery

  List the authenticated user previous generations. Use pagination.nextCursor when pagination.hasMore is true. Retrieve existing outputs before creating unnecessary replacements.

• Get Gallery Stats

  Get generation statistics for the authenticated user ID. Response shape is {status,stats}, not {success,data}. Another user ID is not permitted.

• Get Available Models

  Read current Redis-configured workflows and models; optionally filter by workflowId. Use each model inputFields for public parameter names, supported values, required overrides and defaults. selectionGuidance contains configured strengths/limitations, not guaranteed quality. Choose a compatible model before generation, preserve explicit user choices, and omit settings only when valid defaults are documented. Supplied credentials filter known user restrictions; submission still checks entitlement

• Check Task Status

  Read one owned generation task status. A successful submission is not completed media. Use its task ID to reconcile queued, processing, failed, completed or fallback-pending work; do not resubmit a paid job merely because it is still processing.

• Generate Image To Video

  Animate a supplied starting image into a video clip. The image establishes the starting visual content; it is not merely a style reference. Use reference-to-video for identity or style guidance across multiple references when available. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_status to retrieve the final outcome; do not blindly res

• Generate Text To Video

  Create a video clip from text without a required source image or video. Use image-to-video for a starting frame and reference-to-video for reference-guided generation when available. A multi-shot project may require storyboard planning instead of a single clip. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_status to retrieve the final ou

• Generate Video Ai Edit

  Modify an existing source video according to an instruction while preserving everything not explicitly asked to change. Output duration and aspect ratio follow the source video unless an explicit duration is given. Use image-edit for still images, video-upscale for resolution-only changes, and reference-to-video when the goal is a new video rather than a modified one. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission

• Generate Video Extend

  Continue an existing video past its current ending, using one or more supplied videos as continuation guidance. The video is the source being extended, not a style reference for an unrelated new video — use reference-to-video when only identity or style guidance is wanted. The requested duration is the length of the new continuation, not the combined total. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge

• Generate Video To Video

  Use the existing video-to-video compatibility workflow with supported video or multimodal references. Distinguish source-video editing from reference guidance using the selected model capabilities. Do not assume this workflow extends a clip or that every provider supports base-video editing. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_

• Generate Image Edit

  Modify supplied images according to an instruction. State what must change and what must be preserved. A mask is optional only when the selected model supports it. Do not use for text-only generation or assume pixel-exact preservation. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_status to retrieve the final outcome; do not blindly resu

• Generate Text To Image

  Create a new image from a text prompt. Use image-edit when an existing image must be modified, and an enabled reference-image workflow when a new composition should use reference images. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_status to retrieve the final outcome; do not blindly resubmit after a timeout.

• Generate Text To Music

  Compose original music from a musical brief, with lyrics or image guidance only where supported. Describe mood, instrumentation, tempo, and structure. This does not mix a finished film or guarantee exact synchronization to visual cues. Select an enabled model from get_available_models and use its public parameter names, supported values, and defaults. Submission may charge credits and returns a task ID, not completed media. Use check_task_status to retrieve the final outcome; do not blindly resu

• Publish Artifact

  Save a user-facing UTF-8 file such as SVG, Markdown, JSON or text directly to durable cloud storage, without writing to the agent filesystem. Provide the complete file contents, not a local path or base64. A successful response returns an artifact ID for preview, download and future edits. For a revision, supply the previous artifact ID; originals remain unchanged. Only report the file as saved after this tool succeeds.

• Read Artifact

  Read a previously published UTF-8 artifact by its ID, using the current user permissions. Use its content to edit or reuse a file in a later turn, then publish a new revision with publish_artifact. This does not read local server files.

• Review Candidate

  Submit an independent critic review of an exact candidate in an owned conversation. This executes an LLM review and may incur model cost; it never triggers paid video or other media generation and never authorizes spending. Read get_critic_policy first and honor disabled stages and current limits. A successful submission is not a verdict: retain the returned review.id and use get_critic_review to reconcile preparing, evaluating or unknown work. Supply a caller-generated stable requestId; reuse i

• Get Critic Review

  Read an owned persisted critic review by review ID. Reconciles an in-flight native result without model resubmission or media generation. A preparing, evaluating or unknown review is not a completed verdict. Returns the bounded review, criteria and evidence coverage, not private input snapshots.

• List Critic Reviews

  List persisted critic reviews for one owned conversation. conversationId is required. Defaults to 10 reviews; use smaller pages for detailed findings. Use nextCursor as before for the next page. Large responses retain complete findings in structuredContent but return only metadata in text. Inspect existing reviews after an ambiguous submission outcome rather than automatically resubmitting; this does not run a model or generate media.

• Get Critic Policy

  Read the current critic policy and revision with user authentication. Returns {revision, policy} containing enabled, model IDs, stages, limits, short criterion definitions and energyAxes; excludes administrative instructions and credentials. Both model IDs null means the configured Director default. Honor disabled policy/stages before requesting a review. This read does not run a model, incur review cost, generate media or change policy.