v0.1.29

Changes since v0.1.28:

• fix(lora): make checkpoint resume work end-to-end
• Reject unsupported image and video media roles before a task is queued.
• feat(lora): pick the best checkpoint from a same-condition candidate grid
• fix(lora): train Z-Image and Qwen faces so identity sticks (#76)
• Remove per-image LoRA captions and VLM auto-caption (#81)
• fix(lora): train every image with one caption
• fix(lora): keep face-dataset skip list and identity gates
• fix(lora): harden face-lock training contracts and wizard gates
• feat(frontend): share one model select for badge and description tip (#79)
• Add LTX 2.5 MSR Ref2VA end-to-end with registry-driven ref2va capability.
• Connect multi-reference image edits and upgrade Qwen Image Edit to 2511 (#75)
• fix(lora): train Z-Image and Qwen faces at 1024 with sigmoid noise
• Add optional audio denoise and route H3 Veda SDPA through the shared helper.
• fix(ace_step): VAE tiling, turbo shift, Haar DCW, and cover conditioning parity (#74)
• feat(ace_step): DiT cross-attention lyric timeline + gallery synced scrolling lyrics (#73)
• fix(ace_step): upstream parity fixes for timbre encoder, ResidualFSQ, APG guidance (#72)

v0.1.28

Changes since v0.1.27:

• Wire MiniMax-H3 Ref2VA end-to-end and finish FL2VA accel UI.

v0.1.27

Changes since v0.1.26:

• Fix ACE-Step long-song loops: SFT N-step CFG, no silence tile.
• Add MiniMax-H3 Veda sparse attention and Ref2VA scaffolding.
• Fix ACE-Step cover and DiffRhythm load; add XL Turbo.

v0.1.26

Changes since v0.1.25:

• Fix video upscale on ffmpeg 9 and LTX 2.5 spoken captions.
• Fix DiT stem call dispatch, LTX25 Gemma tokenization, and LTX dialogue prompts.
• Tighten engine extension points around pipeline, hooks, and factories.

v0.1.25

Changes since v0.1.24:

• Fix release CI: pin TypeScript 5.9 for vue-tsc and use Node 22.

v0.1.24

Changes since v0.1.23:

• Raise MLX to 0.32.3, lock MiniMax H3 to full Euler, and bump dependencies (#71)
• LoRA training: face-aware crop, Turbo weight hint, and dataset pre-flight (captions + face audit) (#70)
• Fix Z-Image / Turbo LoRA training failing to retain facial identity (#69)

v0.1.23

Release v0.1.23.

v0.1.22

Release v0.1.22.

v0.1.21

Changes since v0.1.20:

• feat(llm): add backend_llm sidecar and split assistant vs generation cache TTL
• feat(minimax_h3): align DiT with PipeNetwork weights and MiniMax FL2VA aux
• feat(minimax_h3): PipeNetwork parity fixes, perf presets, and Turbo LoRA registry
• docs: add release package install steps for macOS and Linux

v0.1.20

Changes since v0.1.19:

• fix(frontend): correct composerReservePx arity for typecheck