Vector font generation based on cascaded diffusion

Patent metadata

Field	Value
Publication number	US-2025124212-A1
Application number	US-202318507847-A
Country	US
Kind code	A1
Filing date	Nov 13, 2023
Priority date	Oct 11, 2023
Publication date	Apr 17, 2025
Grant date	—

Abstract

In implementation of techniques for vector font generation based on cascaded diffusion, a computing device implements a glyph generation system to receive a sample glyph in a target font and a target glyph identifier. The glyph generation system generates a rasterized glyph in the target font using a raster diffusion model based on the sample glyph and the target glyph identifier, the rasterized glyph having a first level of resolution. The glyph generation system then generates a vector glyph using a vector diffusion model by vectorizing the rasterized glyph, the vector glyph having a second level of resolution different than the first level of resolution. The glyph generation system then displays the vector glyph in a user interface.

First claim

What is claimed is:

1 . A method comprising:
receiving, by a processing device, a sample glyph in a target font and a target glyph identifier; generating, by the processing device using a raster diffusion model, a rasterized glyph in the target font based on the target glyph identifier and the sample glyph, the rasterized glyph having a first level of resolution; generating, by the processing device using a vector diffusion model, a vector glyph by vectorizing the rasterized glyph, the vector glyph having a second level of resolution different than the first level of resolution; and displaying, by the processing device, the vector glyph in a user interface.

2 . The method of claim 1 , wherein the vector glyph includes an ordered sequence of control points using one or more cubic Bezier curve paths based on one or more locations of control points of the rasterized glyph.

3 . The method of claim 1 , wherein the generating the rasterized glyph using the raster diffusion model includes extracting embeddings from the sample glyph in the target font and the target glyph identifier.

4 . The method of claim 3 , wherein the generating of the rasterized glyph by the raster diffusion model is conditioned on the embeddings.

5 . The method of claim 1 , wherein the generating the vector glyph using the vector diffusion model includes synthesizing the vector glyph conditioned on the rasterized glyph.

6 . The method of claim 5 , wherein the generating the vector glyph using the vector diffusion model includes determining a curve of the vector glyph based on the raster glyph by generating a target tensor that represents the curve of the vector glyph in a vector graphics format.

7 . The method of claim 5 , wherein the generating the vector glyph using the vector diffusion model includes upsampling the rasterized glyph.

8 . The method of claim 5 , wherein the generating the vector glyph using the vector diffusion model includes predicting locations for one or more control points that form a curve of the vector glyph.

9 . The method of claim 1 , wherein the rasterized glyph captures a shape and a style of the sample glyph in the target font.

10 . A system comprising:
a memory component; and a processing device coupled to the memory component, the processing device to perform operations comprising:
receiving a glyph in a target font and a target glyph identifier;
generating a rasterized glyph using a raster diffusion model based on the glyph and the target glyph identifier;
generating a vector glyph by vectorizing an encoded location of a control point generated by processing the rasterized glyph using a vector diffusion model; and
displaying the vector glyph in a user interface.

11 . The system of claim 10 , wherein the raster diffusion model is configured to extract embeddings from the glyph and the target glyph identifier.

12 . The system of claim 11 , wherein the raster diffusion model is configured to generate the rasterized glyph as conditioned based on the embedding.

13 . The system of claim 10 , wherein the vector diffusion model is configured to synthesize the vector glyph as conditioned based on the rasterized glyph.

14 . The system of claim 13 , wherein the vector diffusion model is configured to determine a curve of the vector glyph based on the rasterized glyph and generate a target tensor that represents the curve of the vector glyph in a vector graphics format.

15 . The system of claim 13 , wherein the vector diffusion model is configured to upsample the rasterized glyph.

16 . The system of claim 10 , wherein the rasterized glyph captures a shape and a style of the target glyph identifier in the target font.

17 . A non-transitory computer-readable storage medium storing executable instructions, which when executed by a processing device, cause the processing device to perform operations comprising:
receiving a glyph in a target font and a target glyph identifier; generating a rasterized glyph by encoding information from the target glyph identifier and the glyph in the target font in a pixel space using a raster diffusion model; and generating a vector glyph by vectorizing the rasterized glyph using a vector diffusion model.

18 . The non-transitory computer-readable storage medium of claim 17 , wherein the raster diffusion model is configured to extract embeddings from the glyph in the target font and target glyph identifier and generates the rasterized glyph conditioned on the embeddings.

19 . The non-transitory computer-readable storage medium of claim 17 , wherein the vector diffusion model is configured to synthesize the vector glyph conditioned on the rasterized glyph by upsampling the rasterized glyph.

20 . The non-transitory computer-readable storage medium of claim 19 , wherein the vector diffusion model is configured to predict a location for a control point that forms a curve of the vector glyph.

Assignees

• Adobe Inc

Inventors

• Liu Difan
• Fisher Matthew David
• Gharbi Michaël Yanis
• Wang Oliver
• Jacobson Alec Stefan
• Thamizharasan Vikas
• Kalogerakis Evangelos

Classifications

• G06T11/23

  using straight lines or curves · CPC title

• G06T11/40

  Filling planar surfaces by adding surface attributes, e.g. adding colours or textures · CPC title

• G06F40/109 Primary

  Font handling; Temporal or kinetic typography · CPC title

• G06T11/203

  Physics · mapped topic