Skip to content

docs(IndexTTS): describe reference conditioning - #1836

Open
localai-org-maint-bot wants to merge 1 commit into
mudler:mainfrom
localai-org-maint-bot:codex/docs-alignment-20260824
Open

docs(IndexTTS): describe reference conditioning#1836
localai-org-maint-bot wants to merge 1 commit into
mudler:mainfrom
localai-org-maint-bot:codex/docs-alignment-20260824

Conversation

@localai-org-maint-bot

Copy link
Copy Markdown
Collaborator

Row

Documentation alignment for #634 and the conditioning implementation landed by #1790. This is a single-agent, ad hoc documentation repair, so it does not claim a roadmap row.

Before starting

  • Issue/PR search and existing claim: no open PR or issue duplicates this documentation repair; #634 owns the IndexTTS 2.5 feature.
  • Pull request shape selected at row claim: focused documentation correction on a fresh branch from upstream main.
  • Roadmap or matrix row, plus scripts/ready-for-helper.py result when applicable: not applicable; no lifecycle state changes.
  • Exact current-code and test/evidence anchors inspected: src/vllm/model_executor/models/indextts2.cpp, src/vllm/model_executor/models/indextts2_conditioning.cpp, tests/vllm/models/test_indextts2_conditioning.cpp, .agents/specs/indextts-2-5.md, and docs/FEATURES.md.

What changed

The IndexTTS 2.5 model guide and server reference now describe the reference-audio conditioning that the implementation runs. README News records the user-visible change. The text keeps voice-cloning and vLLM-Omni parity explicitly unmeasured.

Evidence

  • Current documentation gates pass. The removed scripts/check-doc-checkpoint.py is not present on main.
  • python3 scripts/check-readme-structure.py: OK.
  • python3 tests/scripts/test_check_readme_structure.py: 19 tests passed.
  • python3 scripts/check-supported-models.py: 41 registered architectures match.
  • python3 scripts/check-site.py: 12 published docs, navigation in bijection.
  • python3 scripts/check-role-discipline.py: OK.
  • python3 tests/scripts/test_check_role_discipline.py: 18 tests passed.
  • python3 scripts/check-agent-record.py: ENGINE=170, MODEL=377, QUANT=84, KERNEL=57, BACKEND=86, ANCHOR-ROT=37.
  • python3 tests/scripts/test_agent_record.py: 109 tests passed.
  • python3 scripts/check-model-checklist.py: checklist matches detailed row states.
  • python3 scripts/check-commit-style.py --range upstream/main..HEAD: OK.
  • python3 scripts/check-commit-trailers.py --range upstream/main..HEAD: OK.
  • Public docs changed only for facts already owned by the landed implementation and feature table.

Speed claims

  • This PR makes NO speed claim.

Honest gaps

No GPU or model run was performed on this host. The PR does not add a new correctness or quality result. It reports the existing deterministic conditioning evidence and keeps oracle parity pending.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true
AI-Assisted: true
Assisted-by: Codex:gpt-5 [Codex]

The model guide and server reference still said that IndexTTS ignored its required reference clip. The implementation now feeds the clip to both conditioning stages, and the feature table already records that behavior.

Keep voice-cloning parity pending because vLLM-Omni is not pinned as an oracle.

FOLLOWING_AGENTS_PROTOCOL

Following-Agents-Protocol: true

AI-Assisted: true

Assisted-by: Codex:gpt-5 [Codex]
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant