0ecb81cf56
The per-family model crates (flux, h3, paint, music, speech, stems, vision, sfx, rife, trellis, beats, common) and the libs/diffusion research binaries read some 190 environment variables that were research knobs: tensor dumps, per-stage timing, oracle-parity and fixture rigs, experiment toggles. The path taken with none of them set is the one that ships; every such knob is deleted with the code it gated, and every losing branch of an experiment toggle goes with its toggle — dead kernels, fields and functions included. What remains are the real configuration variables (the FLUX_*_MODE family, FLUX_GRAPH, the VAE pool cap, the FLUX2 text-encoder residency, H3_VAE_BATCH, the music3 caches and official modes, the stems/beats f16 switches, the weight and data roots) and the build-script variables. Rebased on the runtime cleanup: precision stays explicit everywhere (GemmPrecision, f16_attention_operands, the H3 text precision, DA3's StrictF32 in code); no act16, no H3_ACT_F16, no FLUX_ATTN_F16 or FLUX_VAE_CONV_GEMM reads survive. Reviewed by the delegate reviewer (APPLY, no findings) and gated on the Windows CUDA box: all seventeen model crates check, motion and vision tests, the hub and the diffusion bins — the gate caught one CUDA-only tap marker the Mac never compiles, removed here. On this Mac: the same checks plus motion 24, paint 158 and vision 23 tests. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
73 lines
2 KiB
TOML
73 lines
2 KiB
TOML
[package]
|
|
name = "makepad-ai-speech"
|
|
version = "0.1.0"
|
|
edition = "2021"
|
|
description = "Speech family: IndexTTS-2.5 + Kokoro speech synthesis engines (text in, PCM out). No platform code: OS voices are makepad-system-speech, engine choice is makepad-ai-hub."
|
|
license = "MIT"
|
|
|
|
[features]
|
|
default = ["whisper", "vad", "kokoro", "indextts"]
|
|
# Whisper speech-to-text (CPU SIMD / Metal / CUDA).
|
|
whisper = []
|
|
# Silero voice-activity detection.
|
|
vad = []
|
|
# Kokoro-82M text-to-speech (+ its G2P).
|
|
kokoro = []
|
|
# IndexTTS-2.5 character-voice text-to-speech.
|
|
indextts = []
|
|
# Precompile the Whisper ggml Metal library at build time (else runtime source compile).
|
|
metal-precompile = []
|
|
|
|
[dependencies]
|
|
# THE CUDA store (Whisper's CUDA backend). Must NOT be target-specific: cargo
|
|
# drops target-cfg deps when this crate is a path dep of another standalone
|
|
# `[workspace]`, and the `links = "makepad_ai_cuda"` handshake
|
|
# (DEP_MAKEPAD_AI_CUDA_KERNELS, read in build.rs) only arrives for an
|
|
# unconditional dependency. On macOS/iOS the crate compiles to empty stubs.
|
|
makepad-ai-cuda = { path = "../../cuda" }
|
|
makepad-ai-common = { path = "../common" }
|
|
makepad-ai-loader = { path = "../../loader" }
|
|
makepad-ai-sfx = { path = "../sfx" }
|
|
|
|
[[bin]]
|
|
name = "g2p_test"
|
|
path = "src/bin/g2p_test.rs"
|
|
required-features = ["kokoro"]
|
|
|
|
[[bin]]
|
|
name = "har_bisect"
|
|
path = "src/bin/har_bisect.rs"
|
|
required-features = ["kokoro"]
|
|
|
|
[[bin]]
|
|
name = "kokoro_probe"
|
|
path = "src/bin/kokoro_probe.rs"
|
|
required-features = ["kokoro"]
|
|
|
|
[[bin]]
|
|
name = "parity"
|
|
path = "src/bin/parity.rs"
|
|
required-features = ["kokoro"]
|
|
|
|
[[bin]]
|
|
name = "tts_test"
|
|
path = "src/bin/tts_test.rs"
|
|
required-features = ["kokoro"]
|
|
|
|
[target.'cfg(any(target_os = "macos", target_os = "ios"))'.dependencies]
|
|
makepad-objc-sys = { path = "../../../objc-sys", version = "1.0.0" }
|
|
|
|
[[bin]]
|
|
name = "whisper-test"
|
|
path = "src/bin/whisper_test.rs"
|
|
required-features = ["whisper"]
|
|
|
|
[[bin]]
|
|
name = "vad-test"
|
|
path = "src/bin/vad_test.rs"
|
|
required-features = ["vad"]
|
|
|
|
[[bin]]
|
|
name = "metal-strip-unused"
|
|
path = "src/bin/metal_strip_unused.rs"
|
|
required-features = ["whisper"]
|