AI tools
Generate media and review transcripts, captions, highlights and translated speech.
AI tools use credits for provider work and save downloadable results. Open Mini Apps, search for a tool, and sign in to use it. Your source and text are sent to the selected AI provider for processing. The button shows the charge before you start.
Choose a workflow
| Tool | Result |
|---|---|
| Transcribe recording | Editable transcript, TXT, SRT or VTT |
| Auto captions | Reviewed subtitles or captioned MP4 |
| Isolate speech | Speech-forward MP3 |
| Remove image background | Transparent PNG |
| Upscale image | PNG enlarged 2x or 4x |
| Remove filler words | Reviewed synchronized video and audio cuts |
| Find highlight clips | Up to five reviewed MP4 clips |
| Translate subtitles | Translated TXT, SRT or VTT |
| Generate a voiceover | Narrated MP3 |
| Separate music stems | Six MP3 stems |
| Generate sound effects | A 0.5 to 22 second sound effect |
| Erase an object in a photo | Filled image using your matching mask |
| Extend an image | Generated padding around each edge |
| Replace photo background | A generated background around the detected subject |
| Product scene maker | A product photo in a described scene |
| Relight a photo | A new lighting treatment |
| Restyle a still | A still using a second image as style reference |
| Thumbnail concept generator | Two concepts with an optional title |
| Animate a still | Four seconds of silent video |
| Generate a transition between two stills | Four-second start-to-end image transition |
| Upscale video | Video enlarged 2x or 4x, within the 4K limit |
| Denoise video | Video with AI noise reduction |
| Smooth motion | Video interpolated to 30 or 60 fps |
| Dub a recording | Reviewed translation spoken by one narrator |
| Lip sync to supplied audio | Face video aligned to your audio |
These tools are available in the web catalog. The native mobile catalog currently contains the original seven mini apps.
Generate media
Choose the tool and upload its source files in order. Some tools need two files: an image and matching mask, two stills, or a video and supplied audio. A white mask marks the area to replace. Choose the available settings, enter a description when required, and select Generate. The button shows the credit charge.
Results support image, audio and video previews and individual downloads. Music separation returns vocals, drums, bass, other instruments, guitar and piano from the selected six-stem model. Thumbnail generation makes two artwork concepts; an optional title is placed separately from the generated artwork.
Stills must be PNG, JPEG or WebP and at most four megapixels. The settings panel shows each image's size, and a larger image is flagged before you spend credits. Direct recording tools accept up to 30 seconds. File size is limited to 128 MB. Still animation and two-image transitions generate four silent seconds at 480p. Video enhancement exports H.264, with output capped at 4K. Generated details, product text, faces, lighting and lip motion need review before use.
Extend an image
The preview shows your image inside the frame it grows to, with the new area hatched around it. Added to each edge sets how many pixels appear on every side, between 64 and 512, and the panel shows the result's size. The finished image must stay within four megapixels, so a larger source allows a smaller margin. Describe what lies beyond the edges in the text box above it.
Review transcripts and captions
Transcription and automatic captions accept recordings up to ten minutes. Select the source language and choose Analyze. Correct the returned text and start/end times, then select Export reviewed result. TXT, SRT and VTT exports are supported. MP4 burns captions into video sources; audio-only recordings get a black video canvas with captions.
The review fills the preview area and shows 20 lines per page. Use the bin on a line to omit it, and Previous lines and Next lines for longer transcripts. Cue times must be ordered, nonoverlapping and within the recording. Editing a cue does not spend more AI credits.
Review cuts and highlights
Filler removal identifies candidate word intervals and asks you to review them against the source. Omitting a suggested removal keeps that part of the recording. The export cuts video and audio together. Transcription can miss fillers or place boundaries incorrectly, so listen before accepting a cut.
Highlight selection ranks transcript passages and suggests up to five clips. Review the time ranges, omit unwanted clips, then export individual MP4 files. Each highlight is limited to 60 seconds. Ranking uses the transcript and does not analyze visual scene quality.
Translate subtitles and dub a video
For subtitle translation, paste SRT or VTT with full hour/minute/second timestamps. Choose the target language and Analyze. Translation preserves cue timing; review the wording before exporting a subtitle file.
Dubbing starts from a video and returns translated cues for review. Export uses one selected narrator, fits each generated speech segment to its cue, and replaces the original soundtrack. Voice cloning, speaker matching and background music preservation are not included. Dubbing exports cost 500 credits and support up to 40 cues with 2,000 total characters. If speech cannot fit within the supported tempo range, revise the translation and try again.
Resume and recover
The current tab stores a job receipt so refreshing can resume status polling. Queued jobs can be cancelled with a credit refund. Once a job starts, it runs to completion or failure. Failed jobs return their charged credits. When the file caused the failure, such as an image over four megapixels or a recording that is too long, the message says what to change.
If status cannot be read, keep the receipt and retry after the connection returns. Provider errors do not create placeholder results. A failed or ambiguous paid submission is not automatically submitted again.