Spaces:
Sleeping
Sleeping
|
Download docs/feature-map.md from vikasbanjare/pulse: direct link, hf CLI and curl.
- Browser
- Download file 4.73 kB
-
https://huggingface.co/spaces/vikasbanjare/pulse/resolve/main/docs/feature-map.md
- Command line
-
hf download hf://spaces/vikasbanjare/pulse/docs/feature-map.md
-
curl -L -o feature-map.md https://huggingface.co/spaces/vikasbanjare/pulse/resolve/main/docs/feature-map.md
4.73 kB
A newer version of the Gradio SDK is available: 6.32.0
Feature map: Captions.ai, Kapwing, VEED β Pulse
Scraped 2026-10-17 from the public tool indexes (captions.ai, kapwing.com/tools, veed.io/tools).
Kapwing lists 342 tool pages, VEED 1,161 (most are SEO variants of the same tool). Grouped below by capability,
with how Pulse implements each for free, locally, with no API keys.
| Capability | Captions.ai | Kapwing | VEED | Pulse (free/local) | Status |
|---|---|---|---|---|---|
| Auto captions, word timestamps | β 100+ styles | β subtitles/caption-generator | β add-subtitles, animated-subtitles | faster-whisper large-v3-turbo (CTranslate2, CPU int8), word timestamps + VAD | v1 |
| Caption styles / animations | AI Edit styles (152) | basic presets | basic presets | 20 presets derived from the measured Captions.ai patterns (word-fill, blur-pop, append, line-stack, pills, bars, moving highlight box, hollow inactive, emphasis colour/size/glow) | v1 |
| Caption editor (text, timing, split/merge) | β | β subtitles/editor | β | word list editor, click to delete/emphasise, drag timing on timeline | v1 |
| SRT / VTT export & import | β | β srt/vtt maker, converter | β | /api/subtitles/srt, drop .srt to import | v1 |
| Translate subtitles | β 100+ langs | β subtitles/translator | β | argos-translate offline packages (download on demand) | v2 |
| Text-based editing (delete words = cut video) | β Trim pauses | β smart-cut | β | cuts list + timeline remap; preview skips cuts; export applies | v1 |
| Filler-word removal | β | β smart-cut/filler-words | β | regex over transcript (um, uh, like, you knowβ¦) + one-click delete | v1 |
| Silence removal | β | β audio-editor/silence-remover | β | gaps between words > threshold β cuts | v1 |
| Auto zoom / punch-ins | β AI Edit | β keyframing | β | zoom keys at emphasis cards (pattern G: 1.05β1.08 over 250 ms easeOut) | v1 |
| B-roll | β AI b-roll | β | β ai-b-roll-generator | user-uploaded B-roll clips/images placed on the timeline; no stock API (needs keys) | v2 |
| Text / emoji / sticker overlays | β | β add-text, add-sticker, add-emojis | β | overlay layer (text, native emoji, images) drawn by the same canvas engine | v1 |
| Resize / reframe (9:16, 1:1, 16:9) | β | β resize | β | output aspect + crop position in the compositor | v1 |
| Trim / split / merge clips | β | β | β | timeline cuts; merge = multi-source timeline | v1 / v2 |
| Denoise / enhance speech | β Denoise | β denoise, audio-enhancer | β ai-audio-enhancer | ffmpeg afftdn / loudnorm / compressor chain | v1 |
| Music & SFX | β generate | β add-audio | β | user-provided tracks with ducking (sidechain) ; generation needs a model, out of scope | v2 |
| Text to speech | β | β 100+ langs | β | Kokoro-82M ONNX (free, offline, ~330 MB) | v2 |
| Voice clone / AI twin | β AI twin | β | β ai-clone | needs F5-TTS / XTTS (multi-GB models); exposed as an optional plugin endpoint, off by default | v3 |
| Eye-contact correction | β | β adjust/eye-contact | β | no free local model of acceptable quality; not planned | β |
| AI avatars | β | β | β | out of scope | β |
| Background removal / blur | β | β remove-background | β ai-background-remover | MediaPipe selfie segmentation in browser (free) | v3 |
| Screen / webcam recording | β | β record/* | β | getUserMedia / getDisplayMedia + MediaRecorder β upload | v2 |
| Convert / compress | β | β convert/, compress/ | β | ffmpeg endpoints | v2 |
| Chat / prompt editing | β Text to edit | β ai/video-assistant | β | not planned without an LLM key; rule-based "auto edit" button instead | β |
| Export | β | β | β | client-side WebCodecs (H.264 + AAC via mp4-muxer); server ffmpeg fallback for cuts | v1 |
Captions.ai specifics worth copying
- Style = font + placement + reveal + entrance + emphasis + box. Every one of the 54 premium styles decomposes into those six fields (see
captions-ai-research/caption-motion-patterns.md). - Measured defaults: band centre 67 % of height (9:16), 2 words per chunk, chunk hold 0.5 s median, whole-chunk entrance 80β150 ms blur-pop, inactive words 40 %.
- Four colour slots per basic preset: text, emphasis, stroke/container, glow.
Kapwing specifics
- Tool index categories: Video Editing, Subtitle/Translate/Transcribe, AI Generators, Audio Tools, Text to Speech, Social Media Editing, Convert & Compress, Image & GIF.
- "Smart Cut" = silence + filler removal, the same two operations Pulse derives from the transcript.
VEED specifics
- Same core (subtitles, text-based editing, audio enhance) plus a very long tail of AI image filters; Pulse ignores the image-filter tail.