Bria Ai

Bria — AI Image Generation, Editing & Background Removal

Commercially safe, royalty-free image generation and editing through 20+ API endpoints.

For additional endpoint details, see the Bria API reference for agents.

When to Use This Skill

• Generate images — "create an image of...", "make me a banner", "generate a hero image", "I need a product photo"
• Edit images — "change the background", "make it look like winter", "add a vase to the table", "remove the person"
• Remove/replace backgrounds — "make the background transparent", "cut out the product", "replace with a studio background"
• Product photography — "create a lifestyle shot", "place this product in a kitchen scene", "e-commerce packshot"
• Enhance/transform — "upscale this image", "make it higher resolution", "restyle as oil painting", "change the lighting"

Setup — Authentication

Step 1: Check for existing credentials

bashCopy code
if [ -f ~/.bria/credentials ]; then
  BRIA_ACCESS_TOKEN=$(grep '^access_token=' "$HOME/.bria/credentials" | cut -d= -f2-)
  BRIA_API_KEY=$(grep '^api_token=' "$HOME/.bria/credentials" | cut -d= -f2-)
fi
if [ -z "$BRIA_ACCESS_TOKEN" ]; then
  echo "NO_CREDENTIALS"
elif [ -n "$BRIA_API_KEY" ]; then
  echo "READY"
else
  echo "CREDENTIALS_FOUND"
fi

• READY → skip to making API calls
• CREDENTIALS_FOUND → skip to Step 3
• NO_CREDENTIALS → proceed to Step 2

Step 2: Authenticate via device flow

Source the auth helper and run bria_auth:

bashCopy code
source ~/.agents/skills/bria-ai/references/code-examples/bria_auth.sh
bria_auth

bria_auth will print SIGN_IN_URL=... and USER_CODE=.... Show the user exactly this — nothing more:

Connect your Bria account: Click here to sign in Your code is {USER_CODE} — it's already filled in.

Then wait; bria_auth polls automatically and prints AUTHENTICATED when done.

If it prints an error, the code expired — run bria_auth again.

Do not proceed with any API call until authentication is confirmed.

Step 3: Verify billing status and resolve API key

bashCopy code
source ~/.agents/skills/bria-ai/references/code-examples/bria_auth.sh
bria_introspect

Interpret output:

• BILLING_ERROR: ... — relay the message to the user verbatim and stop. Do not make any API calls.
• TOKEN_EXPIRED — tell the user their session expired and restart from Step 2.
• READY — BRIA_API_KEY is now cached in ~/.bria/credentials. Proceed.

Decision Tree — Which Endpoint to Use

Copy code
Transparent PNG / cutout / remove background?
  → /v2/image/edit/remove_background

Generate image from scratch (text → image)?
  → /v2/image/generate

Edit existing image with text instruction?
  → /v2/image/edit  (use --key images)

Combine 2-4 images — outfit, product, logo, style, or background from one into another?
  → /v2/image/edit  (--key images, then one --image per extra reference)

Remove an object and its shadow or reflection ("remove the bike and its shadow")?
  → /v2/image/edit/erase  ("composite": false, mask only the object)

Change / replace / blur background?
  → /v2/image/edit/replace_background  (prompt: "blur" or describe new bg)

Place product in a lifestyle scene?
  → /v1/product/lifestyle_shot_by_text

Upscale / increase resolution?
  → /v2/image/edit/increase_resolution  (scale: 2 or 4)

Anything else (restyle, relight, reseason, restore, colorize, sketch, blend, outpaint)?
  → See references/capabilities.md for the full endpoint list

How to Call Any Endpoint

bashCopy code
source ~/.agents/skills/bria-ai/references/code-examples/bria_client.sh

# Generate (no image input)
RESULT=$(bria_call /v2/image/generate "" '"prompt": "your description", "aspect_ratio": "16:9", "sync": true')

# Generate with a photographic look (style: kodak_film, hard_flash, hazy_dreamy or rough_realism)
RESULT=$(bria_call /v2/image/generate "" '"prompt": "a woman reading on a city bus", "style": "kodak_film", "seed": 5555, "sync": true')

# Remove background
RESULT=$(bria_call /v2/image/edit/remove_background "/path/to/local/image.png")

# Replace background
RESULT=$(bria_call /v2/image/edit/replace_background "https://example.com/img.jpg" '"prompt": "sunset beach"')

# Edit image (uses images array — pass --key images)
RESULT=$(bria_call /v2/image/edit "/path/to/image.png" --key images '"instruction": "make it look warmer"')

# Edit with reference images — each --image adds the next one, in order
RESULT=$(bria_call /v2/image/edit "https://example.com/man.jpg" --key images \
  --image "https://example.com/santa.png" \
  '"instruction": "dress the man in image 1 in the santa outfit from image 2"')

# Upscale
RESULT=$(bria_call /v2/image/edit/increase_resolution "https://example.com/img.jpg" '"desired_increase": 4')

# Lifestyle shot
RESULT=$(bria_call /v1/product/lifestyle_shot_by_text "/path/to/product.png" '"scene_description": "modern kitchen countertop"')

echo "$RESULT"

Calling convention: bria_call <endpoint> <image_or_empty> [--key <json_key>] [extra JSON fields...]

• Pass a URL, local file path, or "" for endpoints without image input
• Use --key images when the endpoint expects an images array instead of image
• Add --image <url_or_path> once per extra reference image (--key images, up to 4 in total). Order is preserved: the positional image is "image 1", the first --image is "image 2", …
• Returns the result image URL on success, or prints an error to stderr

Editing with several images (2-4): put the image being edited first, references after it, and address them by position — "dress the man in image 1 in the santa outfit from image 2". Say what each reference contributes, in plain prose. Single-image edits need no positional wording.

Generation options: Aspect ratios 1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, plus 16:10, 10:16, 2:1, 1:2, 21:9, 9:21, 3:1, 1:3, 4:1, 1:4, 5:1, 1:5 (not with 4MP, a style_id, or a reference image beside a VGL). Resolution 1MP (default) or 4MP (more detail, +30s). For a photographic look, add "style": kodak_film, hard_flash, hazy_dreamy or rough_realism. It is text-to-image only, so not with style_id, a VGL, reference images, 4MP or lite. Pass "sync": true for a single generated image. Editing endpoints are the other way round — they answer with a status_url you poll, and "sync": true on an edit fails with a gateway timeout.

Advanced: For precise control over generation, use the vgl skill for structured VGL JSON prompts.

Common Failures

• bria_call returns empty / no URL → BRIA_API_KEY was not set. Run Step 3 (bria_introspect) to cache it.
• Async job times out → Some endpoints take 60–90s. If bria_call reports a timeout, retry once; the job may have been queued.
• ERROR 401 → API key is stale. Delete ~/.bria/credentials and re-authenticate from Step 2.
• BILLING_ERROR → Relay message to user verbatim, do not retry API calls.
• Local file not found → Pass the absolute path; bria_client.sh handles base64 encoding automatically.
• /v2/image/edit returns wrong result → Confirm --key images flag is present; this endpoint requires the images array format.

Resources

• Capabilities & Prompt Recipes — Full endpoint table, use-case recipes, and prompt engineering tips
• API Endpoints Reference — Complete parameter documentation for all 20+ endpoints
• Shell Client (bria_client.sh) — bria_call helper: auth, base64, JSON, polling
• Auth Helper (bria_auth.sh) — bria_auth and bria_introspect functions
• Full API docs for agents (llms.txt) — Agent-ready Bria API reference

Related Skills

• vgl — Structured VGL JSON prompts for precise, deterministic control over Fibo image generation
• image-utils — Classic image manipulation (resize, crop, composite, watermarks) for post-processing