---
title: "vuer.workspace.encoders"
section: "Python API"
order: 35
description: "Python API reference for vuer.workspace.encoders"
---

# vuer.workspace.encoders

Image encoding utilities for serving in-memory images via Workspace.

Provides functions to encode numpy arrays or torch tensors as JPEG/PNG bytes
or base64 data URIs for embedding in HTML/JSON.

**Usage**::

    from vuer import Workspace, jpg, png

    ws = Workspace()
    ws.link(lambda: jpg(camera.frame), to="/live/frame.jpg")
    ws.link(lambda: png(depth_map), to="/depth.png")

**Functions**:

- ``jpg(image, quality=90)`` - Encode as JPEG bytes
- ``png(image)`` - Encode as PNG bytes (supports alpha)
- ``b64jpg(image, quality=90)`` - Encode as base64 JPEG data URI
- ``b64png(image)`` - Encode as base64 PNG data URI
- ``decode_b64png(b64)`` - Decode base64 PNG back to PIL Image

**Input format**:

Images should be numpy arrays or torch tensors with values in [0, 1] range.
Shape should be (H, W, C) where C is 1, 3, or 4 channels.

## jpg

```python
def jpg(image: ArrayLike, quality: int=90) -> bytes
```

[Source](https://github.com/vuer-ai/vuer-docs/blob/main/src/vuer/workspace/encoders.py#L59)

Encode image as JPEG bytes.

Does not support transparency - alpha channel is stripped.

:param image: Image array (H, W, C) with values in [0, 1].
             Can be numpy array or torch tensor.
:param quality: JPEG quality (1-100). Default 90.
:return: JPEG-encoded bytes.

Example::

    frame = camera.capture()  # (480, 640, 3) tensor in [0, 1]
    data = jpg(frame, quality=85)

    # Use with Workspace
    ws.link(lambda: jpg(frame), to="/frame.jpg")

## png

```python
def png(image: ArrayLike) -> bytes
```

[Source](https://github.com/vuer-ai/vuer-docs/blob/main/src/vuer/workspace/encoders.py#L90)

Encode image as PNG bytes.

Supports transparency (alpha channel).

:param image: Image array (H, W, C) with values in [0, 1].
             Can be numpy array or torch tensor.
:return: PNG-encoded bytes.

Example::

    rgba = render_with_alpha()  # (480, 640, 4) with alpha
    data = png(rgba)

    # Use with Workspace
    ws.link(lambda: png(depth_colormap), to="/depth.png")

## b64jpg

```python
def b64jpg(image: ArrayLike, quality: int=90) -> str
```

[Source](https://github.com/vuer-ai/vuer-docs/blob/main/src/vuer/workspace/encoders.py#L116)

Encode image as base64 JPEG data URI.

Returns a string that can be used directly in HTML img src
or embedded in JSON responses.

:param image: Image array (H, W, C) with values in [0, 1].
:param quality: JPEG quality (1-100). Default 90.
:return: Data URI string like "data:image/jpeg;base64,..."

Example::

    src = b64jpg(frame)
    # Use in HTML: <img src="{src}">

## b64png

```python
def b64png(image: ArrayLike) -> str
```

[Source](https://github.com/vuer-ai/vuer-docs/blob/main/src/vuer/workspace/encoders.py#L136)

Encode image as base64 PNG data URI.

Returns a string that can be used directly in HTML img src
or embedded in JSON responses. Supports alpha channel.

:param image: Image array (H, W, C) with values in [0, 1].
:return: Data URI string like "data:image/png;base64,..."

Example::

    src = b64png(rgba_image)
    # Use in HTML: <img src="{src}">

## decode_b64png

```python
def decode_b64png(b64: str) -> Image.Image
```

[Source](https://github.com/vuer-ai/vuer-docs/blob/main/src/vuer/workspace/encoders.py#L155)

Decode a base64 PNG data URI back to a PIL Image.

:param b64: Base64 string, with or without data URI prefix.
:return: PIL Image object.

Example::

    img = decode_b64png(data_uri)
    arr = np.array(img) / 255.0  # Back to [0, 1] range