local-ai · review · first seen Oct 24, 01:28 UTC

Ollama adds Phonon voice catalog in release candidate

The Ollama 0.40.11-rc0 release notes list a voice catalog for Phonon and an mlxrunner change that spells out symbols, numbers and links for Phonon. Builders evaluating Phonon should check whether those changes improve their own inputs before treating the release candidate as ready for their workflow.

L0 · primary1 source1 claims checked4 min read

What shipped

The primary artifact is a GitHub release entry titled v0.40.11. Its change list includes “voice: add voice catalog for Phonon” and “mlxrunner: spell out symbols, numbers and links for Phonon.” The entry links those changes to #19242 and #19234, respectively, and labels the full changelog v0.40.10...v0.40.11-rc0. I read this as a release-candidate change, not evidence of a later stable release. The supplied claim specifically confirms the mlxrunner spelling behavior in 0.40.11-rc0; the catalog appears in the accompanying primary-source excerpt. Neither supplies a voice list or a sample output. [1]

The two Phonon entries address different parts of a voice workflow. A catalog concerns which voices a builder can find or select; spelling out symbols, numbers and links concerns how input text is handled by mlxrunner. That distinction matters when testing: discovering a voice does not establish how it will render a link, and a text-handling change does not establish which voices are available. The release entry names both changes, but gives no catalog interface, option names, voice identifiers or instructions for using them. Treat those as questions to answer in the release candidate itself, rather than filling the gaps from the change titles. [1]

The same release entry lists other work, including a qwen3coder parameter change, an OpenCode context change, a vllm-d3 decision change and a llama.cpp version bump. None is described as a prerequisite for Phonon. For this review, the actionable scope is narrower: the named voice catalog and the named mlxrunner handling of symbols, numbers and links. The excerpt does not say whether the changes apply outside mlxrunner, whether existing Phonon behavior is replaced in every case, or whether any application-facing setting has changed. A builder should not infer those details from their presence in one change list. [1]

Why it matters if you build

If a product reads user-supplied text through Phonon, punctuation, numerical strings and URLs are useful test cases because the release candidate explicitly calls out symbols, numbers and links. The likely decision is not simply whether the new output sounds preferable. It is whether spelling those inputs fits the product’s content: a link that should be spoken intelligibly may have different requirements from a passage where a number must remain unambiguous. Compare representative inputs with the behavior your application expects. The source states the intended category of handling, not the exact spoken form of any particular character, number or address. [1]

The catalog is worth inspecting if voice choice is part of your build process, but the release note alone cannot support a migration decision. It does not state how many voices are in the catalog, which voices are included, how a voice is selected, or whether prior selections still behave the same way. That leaves a practical limit on this review: I can identify the new catalog entry, but cannot recommend a specific voice or integration step from the supplied artifact. Test voice discovery separately from speech rendering so that a catalog difference is not mistaken for an mlxrunner difference. [1]

There is no cost, resource-use or performance figure in the supplied material. There is also no quality evaluation, compatibility table or rollout schedule beyond the release-candidate identifier in the changelog. For a builder deciding today, those absences set the size of the commitment: a local evaluation can answer whether the observed behavior suits a particular workflow, while a production-wide change would require evidence this excerpt does not provide. Record the release candidate being tested and the inputs used, so the decision rests on observed behavior rather than on the broad wording of a change title. [1]

How to try it today

1. Find the v0.40.11 release entry and identify the v0.40.11-rc0 changes in its changelog.
2. Inspect the Phonon voice catalog in the release candidate; note what you can actually select and how.
3. Run representative Phonon inputs containing symbols through mlxrunner and record the output.
4. Repeat with numbers and links that occur in your own workflow; compare the result with your required reading.
5. Keep the catalog findings separate from the text-handling findings before deciding whether to adopt the candidate.

What we checked

Every claim below was walked down the source ladder. The rung is how close the source is to the thing itself; the tense is the source's own.

1. shipped L0 · primary

   Ollama 0.40.11-rc0 spells out symbols, numbers and links for Phonon in mlxrunner.

   per GitHub releases · ollama/ollama (L0)

What is not confirmed

The sources do not establish the catalog’s contents or interface; exact pronunciations; behavior outside mlxrunner; compatibility with existing selections; performance or cost effects; or a stable-release date. The release-candidate changelog should not be read as confirmation that those questions are settled. [1]

Who should act, who should wait

Act now if you already evaluate Phonon through mlxrunner and symbols, numbers or links are material to your output: the named change gives you a focused set of inputs to check in 0.40.11-rc0. Inspect the catalog if voice selection is also part of that evaluation. Wait on a broader rollout if you need documented voice choices, exact rendering rules, compatibility evidence or operating costs before changing a shipped workflow. The supplied release entry does not answer those questions. [1]

Sources

• v0.40.11 — GitHub releases · ollama/ollama L0 · primary