Picsha AI

Supported File Formats

Picsha AI's ingestion pipeline is designed to handle a vast array of media types. From raw camera files to heavy design assets, our system automatically generates web-optimized proxies, extracts metadata, and prepares files for seamless delivery.

Ingest Formats

Standard Images

These formats are optimized and converted to next-gen formats (like WebP and AVIF) for high-speed delivery.

  • JPEG / JPG
  • PNG
  • WebP
  • AVIF (AV1 Image File Format)
  • HEIC / HEIF

Complex & RAW Images

Heavy design files and raw camera outputs are automatically processed by our backend cluster (via ImageMagick, LibRaw, and Sharp) into standard web-safe proxy JPEGs or WebPs, ensuring they can be viewed and analyzed by AI.

  • PSD (Adobe Photoshop)
  • AI (Adobe Illustrator)
  • EPS / PostScript
  • TGA (Truevision TGA) & ICO (Windows Icons)
  • RAW Camera Formats: .cr2/.cr3 (Canon), .nef (Nikon), .arw (Sony), .dng (Digital Negative), .raf (Fujifilm), .rw2 (Panasonic), .orf (Olympus)
  • TIFF / TIF
  • BMP
  • SVG (Vector Graphics)

Documents & Sheets

Text is automatically extracted for semantic search, and multi-page documents generate visual poster frames via LibreOffice Headless and Poppler.

  • PDF: Extracts text and creates high-quality preview posters.
  • Word / DOCX / ODT: Extracts text and automatically converts to web-friendly PDFs.
  • Excel / XLSX / XLS / ODS / CSV: Tabular files and spreadsheets are automatically processed for text extraction and visual PDF sheets.
  • PowerPoint / PPTX / ODP: Extracts text from slides and automatically converts to web-friendly PDFs with preview posters.
  • HTML: Standard web pages converted dynamically for archiving and rendering.
  • Text Files: .txt, .md (Markdown), and other plain-text formats.

Video & Audio

Media files undergo specialized processing to enable adaptive streaming, AI transcription, and rich waveform visualizations.

  • Video Formats: (MP4, MOV, WebM, MKV, FLV, MPG, MPEG, OGG/OGV) Automatically generates a thumbnail poster and extracts audio for AI transcription. If the adaptive_stream flag is enabled, videos are additionally processed via AWS MediaConvert for Adaptive HLS streaming.
  • Audio Formats: (MP3, WAV, FLAC, M4A, AAC, OGG, OPUS) Processed via FFmpeg to generate beautiful blue edge-cached wave visualizer representation images (poster.jpg) and transcribed automatically using AI for deep semantic search.

ZIP Archives & Collections

Ideal for bulk uploads from digital media cards or legacy DAM systems.

  • ZIP: Ingested ZIP archives are automatically extracted in the background. Nested photos, videos, and documents are registered as individual assets and enqueued for full metadata extraction, AI analysis, and proxy generation. The original parent ZIP is retained safely in cold storage, but removed from the active UI once extraction completes to keep the workspace clutter-free.

Output Formats

When requesting assets via the Delivery API, Picsha dynamically serves the best format based on the original content type and user request.

Image Outputs

  • Original Format: The unmodified uploaded file (e.g., the original RAW or PSD file).
  • WebP (optimized.webp): Optimized, compressed next-gen format for rapid web delivery.
  • AVIF: High-performance modern web image compression format.
  • JPEG (proxy.jpg or poster.jpg): Web-safe proxies for complex/RAW images, static preview frames for documents and videos, and dynamic waveform visualizations for audio.

Document Outputs

  • Original Format: The unmodified uploaded document.
  • PDF (web.pdf): Dynamically converted PDF versions of non-PDF documents (like Excel or Word sheets) to ensure native web browser previewing.

Video Outputs

  • Original Format: Directly serve the source video file.
  • HLS Stream (playlist.m3u8): Adaptive bitrate streaming playlist generated for seamless playback across different network conditions.
  • JPEG Poster: Static preview frame extracted from the video timeline.

Audio Outputs

  • Original Format: The unmodified audio file.
  • Waveform Poster (poster.jpg): A stunning, high-res 1200x600 visual representation of the audio waves generated automatically.
  • Audio Extract: Extracted audio tracks used internally for transcription.