Sign In

和室 washitsu SD15

629
4.0k
3.1k
45
Verified:
SafeTensor
Type
LoRA
Stats
1,600
2,750
Reviews
Published
Oct 27, 2023
Base Model
SD 1.5
Trigger Words
washitsu, table, scenery, tatami, sliding doors, indoors, book, architecture, cushion, window,
Hash
AutoV2
87D6A6A263

既存のモデルでも「Japanese room」等のプロンプトで和風の部屋を描けたりしますが、違和感が半端ないというか…ちょっと微妙な部屋がでがちです Even with existing models, it is possible to draw Japanese-style rooms with prompts such as "Japanese room," but they tend to look a little too subtle...not quite right.

和室の写真を学習させたらどのくらいディティールがアップするのか? How much detail can be improved by learning pictures of Japanese rooms?

それを検証するために学習を試してみました I tried learning to verify this.

 

prompt:

washitsu, scenery, table, plant, indoors, lamp, sliding doors, tatami, tree, box, book, window, chair, potted plant, vase, television, architecture,

 

prompt:

washitsu, table, scenery, tatami, sliding doors, indoors, book, architecture, cushion, cup, window,

 

田舎の居間っぽく Like the living room of a Japanese country house.

washitsu, scenery, table, indoors, shelf, chair, sunlight, vase, shade, tree, plant, cup, book, lamp, cushion, teapot, plate, cherry blossoms, box, carpet, window, wooden floor, bottle, couch, cabinet, architecture,

 

結果:様々なディティールや日本建築の特徴を描く部分は増すものの、まだ納得できるような描写とは言えない。細部のいい加減さがひどい。もっと多くの資料や資料や素材を集めて細部から丁寧に学習させていく必要がありそう。

Result: Although the portrayal of various details and characteristics of Japanese architecture increases, it is still not a convincing portrayal. The laxity of detail is terrible. It will be necessary to collect more materials, documents and materials to let the students learn carefully from the details.