Reelora API

Type: video — video creation

Photos or video clips + voiceover + karaoke subtitles + music → MP4.

{
  "type": "video",
  "photos": ["upload_id_1", "https://example.com/img.jpg"],
  "text": "Text for voiceover",
  "scene_durations": null, "scene_transitions": null, "target_duration": 30,
  "format": "9_16", "animation": "ken_burns", "transition": "random", "transition_duration": 0.5,
  "tts_provider": "google", "tts_voice": "Chirp3-HD-Aoede",
  "tts_scene": "Speak like a news anchor.",
  "subtitles": true, "sub_style": "yellow_basic", "sub_position": "bottom",
  "bgm_track": "chill_background-lofi-vibes-113884.mp3", "bgm_volume": 0.2
}

audio_mix_type — for clips with original sound

🎬 Dynamic scenes — when the shot changes

• With voice: insert {scene} markers in text between paragraphs. Each segment is timed by the narration and mapped to media in order. Markers are stripped before TTS — the voice never reads them.
• Without voice / explicit: pass scene_durations (e.g. [10, 15, 5]) — one value per media. Video clips are trimmed/looped to fit, photos are shown for the scene length.
• Scenes ↔ media: equal → 1:1; fewer scenes → the last splits across the rest; more → the last media holds the leftover.

✨ Per-cut transitions — how the shot changes

• In text: extend a marker with a type and/or duration: {scene:slide}, {scene:zoom:0.8}, {scene:0.8} (duration only). The transition applies to the cut into that scene.
• Explicit array: scene_transitions (length = photos − 1), e.g. ["slide","zoom","fade"] — one transition per cut. Handy for silent videos with explicit durations.
• Types: fade · dissolve · slide · cover · reveal · zoom · squeeze · wind · pixel · circle · random. Directional ones (slide/cover/reveal/wind) pick a random direction per cut.

🅰️ Text on video — static captions

Add arbitrary captions/titles on top of the video, independent of the subtitles. Works with or without narration. Each caption has its own style, position, timing and fade in/out. Two ways to pass them:

Structured — text_overlays (array, max 10):

"text_overlays": [
  { "text": "BIG TITLE", "style": "bold_yellow", "size": "large",
    "position": "top_center", "start": 0.0, "end": 3.0, "fade": true, "anim": "pop" },
  { "text": "the end", "style": "soft_pink", "position": "bottom_center", "start": 3.0 }
]

Title styles (style): clean_white · bold_yellow · outline_black · soft_pink · elegant_script · news_bar · impact_red.

Codes — text_overlays_raw (string). A universal English syntax (same in every UI language), parsed on the server. One caption per line (or separated by /); fields inside a caption separated by |:

TEXT | attributes | start-end

"text_overlays_raw": "Hello | large italic center | 0-3 / Subscribe | bottom pink pop | 3-end"

• size — small · medium · large
• weight/slant — bold · thin · italic
• vertical — top · middle · bottom; horizontal — left · center · right
• color — white · yellow · red · green · blue · pink · orange · purple · black · cyan
• time — 0-3, 2.5-8, 5-end (sec, dot/comma). No time → caption shows the whole clip
• animation — pop · slide · type (typewriter) · fade · none

⚡ Auto-text — automatic keyword captions

auto_text: true turns on ⚡ Auto-text. Instead of positioning captions yourself, an LLM reads the recognized words of the voice-over, picks the strongest ones (names, numbers, punchy claims) and renders them as animated captions, exactly on the TTS word timestamps. It chooses a modern style, color and appearance animation (pop / slide / typewriter) per word, stacks words spoken together and keeps out of the subtitle zone.

• Requires a voice-over (text non-empty). With no narration there are no word timestamps, so auto_text is ignored.
• Additive: generated captions are appended to your text_overlays / text_overlays_raw, they don't replace them.
• Fully automatic: there are no per-word parameters. If the LLM matches nothing, the render simply proceeds without the extra captions (never fails the request).
• Quota: each successful auto-text render consumes one unit of your monthly allowance (auto_text_limit; free plans limited, paid plans unlimited). Over the limit → 429.
• Auto-text works across every API task type: video, dubbing and subtitles. Translation over the API is done through subtitles with translate_to (there is no separate translation type), so auto_text covers translated subtitles too.

🎬 AI-tables — animated cards over the video

ai_graphics: true turns on 🎬 AI-tables (smart scenes): animated cards drawn over the video — lists, numbers, quotes, comparisons, step lists, bar/percent charts, trend arrows, plus subscribe / like call-to-action cards. Cards avoid faces and your subtitle zone automatically, and are sized for vertical (Reels) playback. Available on video, subtitles (cards in the subtitle language) and dubbing (cards in the dubbing language).

• ai_graphics_mode: auto — an LLM picks the moments from the voice-over (needs narration); semi — auto plus your directives (directives win on overlap); manual — only your directives (works on a silent video).
• ai_graphics_density: 0.6 (rare) · 1.0 (normal) · 1.5 (dense).
• ai_graphics_directives (manual / semi): one card per line — MM:SS[-MM:SS] template args [@position]. Templates: caption, quote, subscribe, like, list, steps, stat, percent, trend, compare. Positions: @center @top @bottom @left @right.
• ai_graphics_sfx: a short sound as each card appears (see Sound effects). null = silent.
• Quota: each successful render consumes one unit of your monthly ai_graphics_limit. Over the limit → 429; plans without the feature → 403. Never fatal: if nothing is selected, the render proceeds without cards.

{
  "type": "video", "photos": ["upload_abc"],
  "text": "Three things to pack for Hawaii...",
  "ai_graphics": true, "ai_graphics_mode": "auto",
  "ai_graphics_density": 1.0, "ai_graphics_sfx": "whoosh"
}

// manual directives example
{
  "ai_graphics": true, "ai_graphics_mode": "manual",
  "ai_graphics_directives": "0:15 subscribe\n0:33-0:45 list What to take | Passport | Money | Tickets\nend like Hit like"
}

🔊 Sound effects

Short sound effects synced to on-screen events. Independent, optional channels — all off by default (null = silent). The value is the filename (without extension) of a sound in the server's SFX library; an unknown value is treated as null.

Sounds are mixed below the voice-over / music, so they accent without masking speech. In the Telegram bot these live inside the ⚡ Auto-text screen, the 🎬 AI-tables screen and the 🎵 Background screen.

Type: transcription — transcription

Audio or video → text + .SRT with timecodes.

{
  "type": "transcription", "file": "upload_id_or_https_url",
  "language": "uk", "ai_prompt": "Medical podcast. Correct terminology.",
  "output_format": "both"
}

// result_urls for transcription
{ "txt_url": "https://cdn.reelora.cc/.../transcript.txt",
  "srt_url": "https://cdn.reelora.cc/.../transcript.srt" }

Type: audio — voiceover (TTS)

Text → MP3 file without video.

{
  "type": "audio", "text": "Hello! This is a Reelora API test.",
  "tts_provider": "gemini", "tts_voice": "Puck",
  "tts_speed": 1.0, "tts_scene": "Warm and friendly tone",
  "tts_style": "expressive"
}

Type: subtitles — subtitles on video

Burns karaoke subtitles into a finished video via Whisper + FFmpeg. Standard+ plan.

{
  "type": "subtitles", "file": "upload_id_or_https_url",
  "language": "uk", "sub_style": "yellow_basic",
  "sub_position": "bottom", "sub_size": "medium"
}

Result: MP4 with subtitles in result_url. Seconds are deducted from the monthly limit.

Type: pause_removal — pause removal

Automatically cuts out silence and pauses from video. Standard+ plan. Silence threshold is detected automatically.

{ "type": "pause_removal", "file": "upload_id_or_https_url" }

Result: processed MP4 in result_url.

Type: dubbing — AI dubbing

Transcribes video, translates, synthesizes a new voice, mixes audio. Three modes. Available on all plans within the monthly limit.

Mode A: auto — fully automatic

{
  "type": "dubbing", "mode": "auto",
  "file": "upload_id_or_https_url",
  "source_lang": "auto", "target_lang": "en",
  "tts_provider": "edge", "tts_voice": "en-US-GuyNeural",
  "mix_mode": "overlay", "overlay_volume": 20,
  "sub_lang": "dubbed", "sub_style": "yellow_basic"
}

Mode B: semi_auto — with verification (2 steps)

// Step 1: transcription
{ "type": "dubbing", "mode": "semi_auto", "step": "transcribe",
  "file": "upload_id_video", "source_lang": "uk", "target_lang": "en" }
// → srt_url: download, edit, upload via /v1/uploads

// Step 2: render with edited SRT
{ "type": "dubbing", "mode": "semi_auto", "step": "render",
  "file": "upload_id_video", "srt_file": "upload_id_of_your_srt",
  "target_lang": "en", "tts_provider": "edge",
  "tts_voice": "en-US-GuyNeural", "mix_mode": "overlay", "overlay_volume": 20 }

Mode C: manual — own SRT file

Skips transcription. You provide your own SRT file containing text in the target language.

{ "type": "dubbing", "mode": "manual",
  "file": "upload_id_video", "srt_file": "upload_id_of_srt_file",
  "target_lang": "uk", "tts_voice": "uk-UA-Wavenet-A", "mix_mode": "replace" }

Type: smart_clip — smart clips

One long video (podcast, interview, vlog) → Whisper detects the language → an LLM picks the strongest self-contained moments → ffmpeg cuts several ready short clips. One call = several clips. Requires a plan with smart clips.

{
  "type": "smart_clip", "file": "upload_id_or_https_url",
  "max_clips": 6, "target_sec": 60, "language": "uk",
  "context": "business podcast, look for tips and strong quotes",
  "reels_mode": true, "subtitles": true, "sub_style": "yellow_basic",
  "auto_text": true, "filters": ["cinema"],
  "webhook_url": "https://mysite.com/hooks/reelora"
}

Clips come back in result_urls.clips[]. result_url = the first clip (for compatibility). Clip fields: title, hook, start/end/duration, score (1–100).

{
  "task_id": 12345, "status": "completed",
  "result_urls": { "clips": [
    { "index": 1, "title": "...", "hook": "...",
      "start": 132.4, "end": 197.8, "duration": 65.4,
      "score": 92, "result_url": "https://cdn.../..._01_....mp4" }
  ] }
}