deAPI platform changelog

New Features

• Expanded the menu with models on the landing page.
• Added the Minimax H3 model for Text to Video and Image to Video.
• Introduced a playground for Audio to Text on the landing page and in the dashboard.

Improvements

• Updated the Enterprise and About Us pages.
• Changed the endpoints of the playgrounds on the landing page and in the dashboard to v2.
• Adjusted error handling after API unification to fix 502 errors.
• Deprecated booster v1, redirecting to v2.

New Features

• Added WhisperLargeV3Ct2, a high-quality speech-to-text model.
• Supports multiple timestamp levels, confidence scores, and optional speaker diarization.
• New output formats include:
  • plain.json: Transcript text without timestamps
  • seg.json: Timestamped segments
  • word.json: Word-level timestamps and confidence scores
  • seg_diar.json: Timestamped segments with speaker labels
  • word_diar.json: Word-level timestamps with speaker labels
  • transcript.txt: Plain readable transcript
• Responses can include detected language, language_probability, timestamp level, segment confidence, and speaker IDs.

YouTube & TikTok Metadata

• Optional source metadata can be requested when transcribing YouTube or TikTok URLs for an additional fee of $0.005 per request.
• Available metadata includes:
  • For TikTok: id, title, description, channel, duration_seconds, view_count, like_count, comment_count, repost_count, save_count, uploader, uploader_url, channel_id, channel_url, webpage_url, timestamp, upload_date, thumbnails[], tags, categories, chapters, heatmap, availability, channel_follower_count
  • For YouTube: <span>id</span>, <span>title</span>, <span>description</span>, <span>channel</span>, <span>duration\_seconds</span>, <span>extractor</span>, <span>view\_count</span>, <span>like\_count</span>

WhisperLargeV3Ct2 is ideal for subtitles, podcasts, interviews, meetings, content indexing, speaker analysis, and social-media transcription pipelines where structured timing and metadata matter.

New Features

• Added a prompt booster that can be enabled in txt/img-to-img requests.
• Introduced multiimage support in the playground dashboard.

Improvements

• Implemented video upscaling in the playground dashboard.
• Added image upscaling in the playground dashboard.
• Added video upscaling on the landing page playground.
• Added image upscaling on the landing page playground.

Bug Fixes

• Fixed section alignment issues on the landing page.
• Resolved various security vulnerabilities.

Performance

• Optimized for SEO and Google AI compatibility.

New Features

• New landing pages for models and model details.
• New landing pages for Enterprise, Terms of Service, Privacy Policy, Comparison and Changelog
• Introduced a Dashboard Playground for text-to-video functionality.