Use high quality reference images
While the Mixfont model will work with any kind of image, stronger reference images will yield better results. To get a better quality font output, you should try to isolate the letterforms and make sure they have clean edges, visible shapes, and enough letters to communicate the style. The more evidence the model can gather from an input image, the better the results will be. With input images that display multiple font styles, the model will usually choose the style that is larger or takes up more area in the image. It is recommended that you first crop the images to the exact text region before submitting it to the model for best results.| Input | Output |
|---|---|
![]() | Canyon-Blob-Display.ttf |
![]() | Gossamer-Editorial-Serif.ttf |
![]() | Astral-Groove-Display.ttf |
rounded, italicized modern sans serif font | Borealis-Soft-Italic-Glow.ttf |
Choose the right input format
Start with the input that best captures the style you want. Use text generation when you can describe the type direction. Use image generation when a visual reference is the clearest source of truth.| Goal | Recommended input | Best practice |
|---|---|---|
| Explore many style directions | Text prompt | Vary one style detail at a time so you can compare results clearly. |
| Match a sketch, sign, logo, or sample | Image URL | Use a clear image with readable letterforms and strong contrast. |
| Build prompt presets into a product | Text prompt | Store reusable prompt templates for common categories and moods. |
| Recreate an existing visual direction | Image URL | Use the reference image as the main style source, then test the TTF in real layouts. |
Write specific prompts
Good prompts describe the type category, visual style, intended use case, and distinctive details. Avoid vague prompts that only describe mood.| Stronger prompt | Weaker prompt |
|---|---|
A high-contrast editorial serif with sharp wedge serifs for fashion headlines | A stylish fashion font |
A soft geometric sans with rounded terminals for a wellness app identity | A calm modern font |
A bold compressed grotesk with industrial spacing for concert posters | A cool poster font |
Pick the right glyph set
Usestandard when English support is enough for exploration, prototypes, headings, logos, and first drafts. It generates 72 glyphs for English with basic letters, numbers, and punctuation.
Use extended when you need production candidates that support Latin-language text beyond English. It generates 319 glyphs for all Latin languages, including special characters, costs more credits, and may take 2-3 minutes to complete.
| Glyph set | Best for | Glyphs | Credits |
|---|---|---|---|
standard | English concepting, prototypes, and exploration | 72 | 20 |
extended | Production candidates for all Latin languages | 319 | 50 |
standard if you only need coverage for English. Standard inference runs are much faster and will be enough for most workflows. Choose extended if you want a more complete font with broader character coverage.
Handle jobs asynchronously
Font generation runs as an asynchronous job. Store the returned generationid and poll_url, then poll until the job reaches a terminal status.
| Status | Meaning |
|---|---|
preparing | Mixfont is preparing the request for generation. |
queued | The job is waiting to run. |
running | The model is generating the font. |
succeeded | The generated TTF is ready in ttf_url. |
failed | The job could not complete. Check error. |
cancelled | The job was cancelled before completion. |
succeeded, failed, and cancelled as terminal states. Add a timeout in your application so you do not poll indefinitely.
Use the client libraries when you want Mixfont to handle polling. Use the REST API when you need direct control over job creation, status checks, retries, and storage.
Store results for review
When a job succeeds, download or persist the generatedttf_url before you show it as a finished asset. Keep the generation id, original prompt or image URL, and glyph_set with the result so your team can compare outputs later.
Test generated fonts in real content, not only alphabet previews. Check headings, numbers, punctuation, short labels, and the longest strings your design needs to support.
Rehost generated font files
Returned TTF files are temporary and will be deleted within 24 hours. Download each generated TTF file after the job succeeds, then rehost it on your own storage before you use it in production. Update your@font-face declarations, CDN references, and saved asset records to use your rehosted file URL instead of the temporary ttf_url.
Plan for credits and errors
Use the pricing table to estimate generation costs from each request’sglyph_set. Make sure your workspace has enough credits before you submit a batch of generations.
If the API returns a 402 error, prompt the user to add credits before retrying. If a generation returns failed, show the error value and let the user adjust the prompt, image, or glyph set.
Protect your API key
Call the Mixfont API from server-side code when possible. Do not expose your API key in browser JavaScript or public client bundles. Read the API key from an environment variable, such asMIXFONT_API_KEY, and send it with the x-api-key header or through a Mixfont client library.
Font generation guide
Build the full generation workflow with JavaScript, Python, or cURL.
API reference
Review request and response fields for text generation.




