feat(eris): supertonic text-to-speech, ten swedish voices at 44 kHz - #1782
Merged
Merged
Conversation
Contributor
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
piper has three Swedish voices and all of them are
medium- there is nohightier for the language, so it has no headroom left. Supertonic 3 is four small ONNX graphs rather than an autoregressive model, which is what keeps it fast without a GPU.Measured on the same sentence, synthesis seconds per second of audio:
Supertonic is faster than what it replaces, at double the sample rate - though piper was measured on icarus and Supertonic on eris, so that comparison is not like-for-like. Chatterbox was the other candidate and is out: RTF 1.8 on CPU, and its GPU path is blocked by nixpkgs' onnxruntime advertising MIGraphX while omitting the library.
Three pieces, same shape as the ASR work:
packages/supertonic- the SDK from PyPI, pure Python, all four dependencies already in nixpkgspackages/wyoming-supertonic- Supertonic has no server of its own, so Home Assistant cannot reach it otherwise. Converts the SDK's float32 output to the interleaved 16-bit PCM wyoming expects, chunks it so playback can start before a long sentence finishes, and loads the model once at startup rather than per request.profiles/supertonic-tts.nix- one unit, weights persisted under/keep, with theRequiresMountsForfix from fix(qwen-asr): wait for the state mount, and give llama-server a HOME #1781 applied from the start.eris enables it on F5.
Verified over the protocol before pushing: all ten voices advertised through
Describe, and synthesis at RTF 0.187-0.228 at 44.1 kHz.🤖 Generated with Claude Code
https://claude.ai/code/session_01XfYTVcQVjCevVeoJofu8FF