Add AI Enhance toggle for better genre accuracy

When enabled, uses the LLM to enrich genre/style tags into detailed
music descriptions and generate proper BPM, key, and time signature
metadata (CoT features). Fixes genre tags like "pop, rock" producing
ballad-like output by matching Gradio UI default behavior.

- Add enhance toggle in Style of Music card header
- Gate CoT metas/caption/language by enhance OR thinking flag
- Remove unsupported --lm-backend/--lm-model from Python fallback
- Add i18n translations (en/zh/ja/ko) with tooltip
This commit is contained in:
fspecii
2026-02-10 13:24:36 +02:00
parent ac99e9efcb
commit 8433349af5
4 changed files with 43 additions and 14 deletions
+12
View File
@@ -265,6 +265,9 @@ export const translations = {
randomSeedRecommended: 'Randomized every run (recommended)',
fixedSeedReproducible: 'Fixed seed for reproducible results',
enterFixedSeed: 'Enter fixed seed',
enhance: 'AI Enhance',
enhanceHint: 'better genre accuracy, slightly slower',
enhanceTooltip: 'Uses the AI language model to enrich your genre/style tags into a detailed music description and generate proper BPM, key, and time signature. Improves genre accuracy but adds 10-20s to generation time. No extra VRAM needed.',
thinkingCot: 'Thinking (CoT)',
mp3Smaller: 'MP3 (smaller)',
flacLossless: 'FLAC (lossless)',
@@ -861,6 +864,9 @@ export const translations = {
randomSeedRecommended: '每次运行随机(推荐)',
fixedSeedReproducible: '固定种子以获得可重现结果',
enterFixedSeed: '输入固定种子',
enhance: 'AI 增强',
enhanceHint: '更准确的风格,稍慢',
enhanceTooltip: '使用AI语言模型将风格标签丰富为详细的音乐描述,并生成准确的BPM、调性和拍号。提高风格准确度,但增加10-20秒生成时间。无需额外显存。',
thinkingCot: '思考模式(CoT',
mp3Smaller: 'MP3(较小)',
flacLossless: 'FLAC(无损)',
@@ -1457,6 +1463,9 @@ export const translations = {
randomSeedRecommended: '毎回ランダム化(推奨)',
fixedSeedReproducible: '再現可能な結果のための固定シード',
enterFixedSeed: '固定シードを入力',
enhance: 'AI エンハンス',
enhanceHint: 'ジャンル精度向上、やや遅い',
enhanceTooltip: 'AIがジャンル/スタイルタグを詳細な音楽説明に変換し、適切なBPM、キー、拍子を生成します。ジャンルの精度が向上しますが、生成に10-20秒追加されます。追加VRAMは不要です。',
thinkingCot: '思考(CoT',
mp3Smaller: 'MP3(小さい)',
flacLossless: 'FLAC(ロスレス)',
@@ -2053,6 +2062,9 @@ export const translations = {
randomSeedRecommended: '매 실행마다 무작위 (권장)',
fixedSeedReproducible: '재현 가능한 결과를 위한 고정 시드',
enterFixedSeed: '고정 시드 입력',
enhance: 'AI 향상',
enhanceHint: '더 정확한 장르, 약간 느림',
enhanceTooltip: 'AI 언어 모델을 사용하여 장르/스타일 태그를 상세한 음악 설명으로 변환하고 적절한 BPM, 키, 박자를 생성합니다. 장르 정확도가 향상되지만 생성 시간이 10-20초 추가됩니다. 추가 VRAM이 필요하지 않습니다.',
thinkingCot: '생각 (CoT)',
mp3Smaller: 'MP3 (작음)',
flacLossless: 'FLAC (무손실)',