Skip to content

[feat] Expose H3 duration and resolution choices in model metadata - #2085

Draft
aryan5v wants to merge 1 commit into
mainfrom
aryan/serve-generation-capabilities
Draft

aryan5v wants to merge 1 commit into
mainfrom
aryan/serve-generation-capabilities

Conversation

@aryan5v

@aryan5v aryan5v commented Nov 10, 2026 •

Copy link
Copy Markdown
Collaborator

Purpose

Clients currently see only a served alias from /v1/model_info, so an initial five-second 480p serve default can become an accidental permanent UI limit. Publish known MiniMax-H3/FastH3/Omni and Pika H3 Flex output geometry separately from request defaults.

Changes

  • Add versioned video_output_capabilities: the 24 fps causal-VAE frame grid, integer durations 5–15 seconds, and recommended 480p/768p canvases across aspect families.
  • Reuse family geometry rules; unknown families return null. Explicit MiniMax-H3 recommendations stay within the adapter's canvas area limit after lattice rounding.
  • Keep model support separate from device capacity (hardware_capacity_verified: false). This does not change generation, trained sampling schedules, attention, residency, or hardware policies.

Test Plan

  • Isolated CPU-only pytest run of fastvideo/tests/entrypoints/test_generation_capabilities.py: 4 passed, covering both families, frame alignment, canvas validity, unknown families and the mounted HTTP route. Used existing Linux dependencies; no model loading or GPU generation.
  • pre-commit run --files docs/design/server_contracts/openai.md fastvideo/entrypoints/openai/common_api.py fastvideo/entrypoints/openai/generation_capabilities.py fastvideo/tests/entrypoints/test_generation_capabilities.py: all applicable checks passed, including mypy.

Draft: this is geometry discovery, not evidence that 768p/15-second reference-heavy generations fit any particular GPU. The companion opt-in harness reader is dreamverse-harness #139. No live worker was patched or restarted to validate this change.

@mergify

mergify Bot commented Nov 10, 2026

Copy link
Copy Markdown
Contributor

Merge Protections

🔴 1 of 1 protections blocking · waiting on 👀 reviews and 🤖 CI

Protection Waiting on
🔴 PR merge requirements 👀 reviews and 🤖 CI

🔴 PR merge requirements

Waiting for

  • #approved-reviews-by>=1
  • check-success=fastcheck-passed
  • check-success=full-suite-passed
  • check-success~=pre-commit
This rule is failing.
  • #approved-reviews-by>=1
  • check-success=fastcheck-passed
  • check-success=full-suite-passed
  • check-success~=pre-commit
  • title~=(?i)^\[(feat|feature|bugfix|fix|refactor|perf|ci|doc|docs|misc|chore|kernel|new.?model|skill|skills|infra)\]

@mergify mergify Bot added type: feat New feature or capability scope: inference Inference pipeline, serving, CLI scope: infra CI, tests, Docker, build scope: docs Documentation labels Nov 10, 2026

This branch has not been deployed

No deployments
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

scope: docs Documentation scope: inference Inference pipeline, serving, CLI scope: infra CI, tests, Docker, build type: feat New feature or capability

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant