Pollinations.ai API for AI generation and analysis - text, images, videos, audio, vision, and transcription. Use when user requests AI-powered content (text completion, image generation/editing, video generation, audio/TTS, image/video analysis, audio transcription) or mentions Pollinations. Supports 25+ models with OpenAI-compatible endpoints.
Pollinations v1.1.1
Unified AI platform for generating and analyzing text, images, videos, and audio with 250+ text, 75+ image, 25 video and 40 audio models.
API Key
Get free or paid keys at https://enter.pollinations.ai
- Secret Keys (
sk_): Server-side, no rate limits (recommended) - Optional for many operations (free tier available)
Runtime Requirements
| Type | Name | Required |
|---|---|---|
| Env | POLLINATIONS_API_KEY | Optional (free tier works without) |
| Bin | curl | Yes |
| Bin | jq | Yes |
| Bin | base64 | Yes |
Operations & Scripts
1. Text/Chat Generation (scripts/chat.sh)
Generate text using 250+ LLM models with OpenAI-compatible API.
Usage:
scripts/chat.sh "your message"
scripts/chat.sh "your message" --model claude --temp 0.7
scripts/chat.sh "explain quantum physics" --model openai --max-tokens 500
scripts/chat.sh "list 3 colors" --json --model openai
scripts/chat.sh "solve this step by step" --model openai-reasoning --reasoning-effort high
scripts/chat.sh "translate to French" --system "You are a translator" --model gemini
Options:
--model MODEL— Model name (default: openai)--temp N— Temperature 0-2 (default: 1)--max-tokens N— Max response length--top-p N— Nucleus sampling 0-1--seed N— Reproducibility (-1 = random)--system "PROMPT"— System prompt--json— Force structured JSON response--reasoning-effort LVL— For reasoning models (e.g. openai-reasoning, grok-reasoning, kimi-thinking): high/medium/low/minimal/none--thinking-budget N— Token budget for reasoning models
Models: short aliases such as openai (free, default), deepseek, grok, kimi, llama, minimax (free) and claude, gemini, gemini-large, gemini-search, mistral, perplexity (paid), or a full id like openai/gpt-5.5. Ids are now provider/name; old aliases may change, so check scripts/models.sh text [--free] (shows aliases and free/paid).
Simple text (no script needed):
curl "https://gen.pollinations.ai/text/Hello%20world"
2. Image Generation (scripts/image.sh)
Generate images from text prompts with multiple models and options.
Usage:
scripts/image.sh "a sunset over mountains"
scripts/image.sh "a portrait" --model flux --width 1024 --height 1024
scripts/image.sh "logo design" --model gptimage --quality hd --transparent
scripts/image.sh "photo" --enhance --nologo --private
scripts/image.sh "art" --negative "blurry, low quality" --seed 42
Options:
--model MODEL— Model (default: flux)--width N— Width 16-2048px (default: 1024)--height N— Height 16-2048px (default: 1024)--seed N— Reproducibility--output FILE— Output filename--enhance— AI prompt improvement--negative "TEXT"— Negative prompt (what to avoid)--nologo— Remove watermark--private— Private generation--safe— Enable NSFW filter--quality LEVEL— low/medium/high/hd (gptimage only)--transparent— Transparent background PNG (gptimage only)--image-url URL— Source image for image-to-image
Models: free: flux (default), kontext, klein, z-image; paid: gptimage, seedream, nanobanana, nanobanana-pro, flux-2-pro, and more. Use scripts/models.sh image [--free] to list all.
3. Image Editing / Image-to-Image (scripts/image-edit.sh)
Transform or edit existing images using AI.
Usage:
scripts/image-edit.sh "make it blue" --source "https://example.com/photo.jpg"
scripts/image-edit.sh "add sunglasses" --source photo.jpg --model kontext
scripts/image-edit.sh "convert to watercolor" --source input.png --output watercolor.jpg
Options:
--source URL/FILE— Source image (URL or local file, required)--model MODEL— Model (default: kontext)--seed N— Reproducibility--negative "TEXT"— Negative prompt--output FILE— Output filename
4. Video Generation (scripts/image.sh with video models)
Generate videos from text prompts or images.
Usage:
scripts/image.sh "a cat playing piano" --model veo --duration 6
scripts/image.sh "ocean waves" --model seedance-2.0 --duration 8 --aspect-ratio 16:9
scripts/image.sh "timelapse" --model veo --duration 4 --audio
scripts/image.sh "animate this" --model seedance-2.0 --image-url "https://example.com/photo.jpg"
Options (in addition to image options):
--model MODEL— Video model (required)--duration N— Length in seconds (veo: 4/6/8; others vary, see the model)--aspect-ratio RATIO— 16:9 or 9:16--audio— Enable audio generation (veo only)--image-url URL— Source image for image-to-video
Frame interpolation (veo): Pass two images for first/last frame interpolation using the API directly:
https://gen.pollinations.ai/image/prompt?model=veo&image[0]=first_frame_url&image[1]=last_frame_url
Models: all video models are paid (need a funded key): veo (4-8s, audio, frame interpolation), seedance-2.0 / seedance-2.5, wan / wan-2.7, kwaivgi/kling-v3.0-std, grok-imagine-video, happyhorse. Run scripts/models.sh video.
5. Text-to-Speech / Audio (scripts/tts.sh)
Convert text to speech via the OpenAI-compatible /v1/audio/speech endpoint.
Usage:
scripts/tts.sh "Hello world"
scripts/tts.sh "Bonjour le monde" --voice nova --format mp3
scripts/tts.sh "Welcome" --model eleven --voice rachel --format wav --output welcome.wav
scripts/tts.sh "Speak warmly" --model qwen-tts-instruct --instruct "soft and gentle"
scripts/tts.sh "upbeat jazz" --model elevenmusic --output music.mp3
Options:
--voice VOICE— Voice (default: nova)--format FORMAT— Output format (default: mp3)--model MODEL— Model (default: tts-1)--speed N— Speech speed 0.25–4.0--seed N— Deterministic output--instruct TEXT— Emotion/style hint (qwen-tts-instruct only)--output FILE— Output filename
Voices: depend on the model. tts-1 / tts-1-hd (default, free): alloy, echo, fable, onyx, nova, shimmer. ElevenLabs models (--model eleven): 35 voices such as rachel, domi, bella, adam, antoni, josh, sam, daniel, george, brian (or a custom voice UUID). Other models (kokoro, qwen3-tts, grok-tts, gemini TTS, mai-voice) have their own lists: scripts/models.sh audio --json | jq shows the model and its voices.
Formats (6): mp3, opus, aac, flac, wav, pcm
Models: tts-1 (default, free), tts-1-hd (free), eleven (ElevenLabs v3, paid), qwen-tts-instruct (instructable speech, paid), kokoro (paid), music (ElevenLabs music, paid). Run scripts/models.sh audio [--free].
6. Image Analysis / Vision (scripts/analyze-image.sh)
Analyze and describe images using vision-capable AI models.
Usage:
scripts/analyze-image.sh "https://example.com/photo.jpg"
scripts/analyze-image.sh photo.jpg --prompt "What objects are in this image?"
scripts/analyze-image.sh image.png --model claude --prompt "Extract all text from this image"
Options:
--prompt "TEXT"— Analysis question (default: "Describe this image in detail")--model MODEL— Vision model (default: openai, free)
Input: URL or local file (jpg, png, gif, webp)
Models: any text model that accepts images: openai (free), grok, kimi (free); gemini, gemini-large, claude (paid). Use scripts/models.sh vision [--free].
7. Video Analysis (scripts/analyze-video.sh)
Analyze video content using AI vision models.
Usage:
scripts/analyze-video.sh "https://example.com/video.mp4"
scripts/analyze-video.sh recording.mp4 --prompt "Summarize the key moments"
scripts/analyze-video.sh clip.mov --model gemini-large --prompt "Count the people"
Options:
--prompt "TEXT"— Analysis question (default: "Describe this video in detail")--model MODEL— Video-capable model (default: gemini)
Input: URL or local file (mp4, mov, avi)
Models: needs a model with video input: gemini (default, paid), gemini-large; community models (see scripts/models.sh text) with video input may be free but unreliable.
8. Audio Transcription (scripts/transcribe.sh)
Transcribe audio files to text via the Whisper-compatible /v1/audio/transcriptions endpoint.
Usage:
scripts/transcribe.sh recording.mp3
scripts/transcribe.sh podcast.wav --model scribe --language en
scripts/transcribe.sh "https://example.com/audio.mp3" --language fr --format srt
scripts/transcribe.sh meeting.m4a --format verbose_json --temperature 0
Options:
--model MODEL— whisper-large-v3 (default), whisper-1, scribe (ElevenLabs)--language CODE— ISO-639-1 hint (e.g. en, fr, ja) for better accuracy--prompt TEXT— Optional context/style prompt--format FORMAT— json (default), text, srt, vtt, verbose_json--temperature N— Sampling temperature 0–1
Input: Local file or URL (mp3, mp4, mpeg, mpga, m4a, wav, webm)
9. List Available Models (scripts/models.sh)
Dynamically list all available models from the API.
Usage:
scripts/models.sh # List all models: category, id, [aliases], free|paid
scripts/models.sh text --free # Only free models (any category; add --json for raw data)
scripts/models.sh text # Text/chat models only
scripts/models.sh image # Image generation models
scripts/models.sh video # Video generation models
scripts/models.sh vision # Vision/analysis models
scripts/models.sh audio # Audio/TTS models
API Endpoints Reference
| Operation | Endpoint | Method |
|---|---|---|
| Simple Text | /text/{prompt} | GET |
| Chat Completion | /v1/chat/completions | POST |
| Image Generation | /image/{prompt}?{params} | GET |
| Image-to-Image | /image/{prompt}?image={url}&{params} | GET |
| Video Generation | /video/{prompt}?{params} | GET |
| Image Analysis | /v1/chat/completions (with image_url) | POST |
| Video Analysis | /v1/chat/completions (with video_url) | POST |
| Text-to-Speech | /v1/audio/speech | POST |
| Audio Transcription | /v1/audio/transcriptions (multipart) | POST |
| Media Upload | https://media.pollinations.ai/upload | POST |
| List Text Models | /v1/models | GET |
| List Image Models | /image/models | GET |
| List Audio Models | /audio/models | GET |
| List Vision Models | /text/models | GET |
Tips
- Free tier available: Many operations work without an API key (rate limited)
- OpenAI-compatible: Chat endpoint works with existing OpenAI integrations
- Reproducibility: Use
seedparameter for consistent outputs across all operations - Image enhancement: Use
--enhancefor AI-improved prompts on image generation - JSON mode: Use
--jsonflag on chat for structured data extraction - Reasoning models: Use
--reasoning-effortwith reasoning models (openai-reasoning, grok-reasoning, kimi-thinking) for controlled thinking depth - Video from images: Use
--image-urlwith seedance-2.0 for image-to-video, or veo frame interpolation - Audio on video: Use
--audiowith veo model for videos with sound - Local files: Analysis scripts (image, video, transcription) accept both URLs and local files
- Data leaves your machine: prompts, images, video and audio are sent to gen.pollinations.ai (and local images for editing are uploaded to media.pollinations.ai). Selecting a model such as claude, gemini or openai routes the content to that vendor through Pollinations. Do not send anything you would not share with a third party.
- Transcription inputs: only local files and http(s) URLs are accepted; other URL schemes (file://, ftp://) are rejected.
- Private mode: Use
--privateto keep generations off public feed
API Documentation
Full docs: https://enter.pollinations.ai/api/docs
Related skills
- Downloads
- 3.5kAll time
- Security audit
- Pass
- Last updated
- 1w ago
- Current version
- v1.1.1
- License
- MIT-0
