Plan the shootNew
2 toolsPre-production, storyboard, scene planning and more.
Describe it. Agent P builds it.
P Agent P Beta
@Image1@Image2@Image3
Create a short 3-scene cinematic film called “Last Train Home”. Scene 1: a young woman in a red coat waits alone on an empty subway platform at night. Scene 2: the train arrives and she steps aboard. Scene 3: she sits by the window watching the city lights blur past. Keep the SAME character throughout.
- Scenes3 × 5 seconds
- Runtime12 seconds — the fades overlap
- CharacterSame woman in all three scenes
- TransitionsFades between scenes
- SoundtrackMoody cinematic orchestral
- Generated three scene images, one character throughout 1@Image1@Image2@Image3
- Animated each still into a 5-second clip 2@Video1@Video2@Video3
- Generated a moody orchestral soundtrack 3@Audio1
- Composed the three clips with fade transitions 4
- Mixed the music under the finished film 5@Video5
Working out the next step…
5 steps
#1 Images 3 files
@Image1
@Image2
@Image3
#2 Clips 3 files
@Video1
@Video2
@Video3
#3 Soundtrack 1 file
@Audio1
#4 Final film 1 file
@Video5
Assets land here as each step finishes.
P Agent P Beta
the-last-light.txt
“The Last Light” — every evening an old man walks to the same hill with a lantern. A girl follows and asks why. Someone once kept a light on for him. One evening the wind takes his flame and she relights it. Soon the whole hill is lanterns.
- Scenes5 × 5.04 seconds
- Ratio16:9 at 1080p
- Runtime19 seconds — the fades overlap
- Score30s orchestral, 60 BPM
- NarrationOne closing line, voiced
- Generated five 5-second scenes straight from the script 1@Video1@Video2@Video3@Video4@Video5
- Stitched them into one cut with 1.5s crossfades 2@Video6
- Scored it with a 30-second orchestral track 3@Audio1
- Voiced the closing line 4@Audio2
- Assembled the score and the narration into a single audio bed 5
- Muxed that bed onto the cut 6@Video7
Working out the next step…
6 steps
#1 Scenes 5 files
@Video1
@Video2
@Video3
@Video4
@Video5
#2 Cut 1 file
@Video6
#3 Score 1 file
@Audio1
#4 Narration 1 file
@Audio2
#5 Final film 1 file
@Video7
Assets land here as each step finishes.
P Agent P Beta
Make me a 5 second video of a red sports car driving through rain at night.
- Ratio16:9
- Resolution1080p
- Duration5.04s measured
- AudioNone — Lite mode has no audio track
- Wrote the full shot description from one line of brief 1@Video1
- Generated a 5-second 1080p clip 2
- Reported the measured length, and that Lite mode delivered no audio 3
Working out the next step…
3 steps
#1 Clip 1 file
@Video1
Assets land here as each step finishes.
P Agent P Beta
aurora-line.txt
Voice this ad line, dub it into Spanish, put the two takes together, and give me a timed SRT for the English one.
- LineOne sentence
- Voicesloongjohn · Kore
- LanguagesEnglish · Spanish
- Caption formatSRT
- Voiced the line in English 1@Audio1
- Dubbed it into Spanish 2@Audio2
- Rendered the two takes as one dialogue track 3@Audio3
- Transcribed the English take to a timed SRT 4@File1
Working out the next step…
4 steps
#1 Voiceover 1 file
@Audio1
#2 Spanish dub 1 file
@Audio2
#3 Dialogue 1 file
@Audio3
#4 Captions 1 file
@File1
Assets land here as each step finishes.
Free trial · Cancel anytime
Trusted by creators and businesses
The last mile
Everyone else stops at the pretty picture.
Generate, cut out, upscale, trace to SVG, check against A4 at 300 DPI, export. Six steps, six tabs — and the printer still rejects it. Agent P does all six in one thread.
The way it works today
- 01Generate the design in an image tool
- 02Upload elsewhere to cut the background out
- 03A third tool to upscale it for print
- 04Illustrator to Image Trace it to vector
- 05Set the artboard, paper size and DPI by hand
- 06Export each format separately
Six subscriptions. Six chances to get the spec wrong.
With Agent P
- 01Describe the job in plain English, once
- 02Agent P picks the tools and runs them in order
- 03Each result feeds straight into the next step
- 04Reuse any output by name — @Image1
- 05Named formats, not approximations
- 06Every file lands in the same thread
One thread. 22 credits for the finished listing.
Spec-correct
Named formats, not approximations.
SVG for the cutter. 300 DPI for the press. Transparent PNG for the mockup. 9:16 for Reels. SRT for the player. MP4 for the ad platform. Agent P checks the file against the spec you named before it hands it back.
Ask for the spec by name
- Vectorize to SVG — clean paths that scale to any size
- Prepare for print — a real paper size, at a real DPI
- Remove background — transparent, no green screen
- Crop or reframe — 9:16, 1:1, 16:9
- Convert subtitles — SCC, SRT, VTT, ASS
- Convert a video — MP4, MOV, MKV, AVI, WMV
Six of the 45 tools. All of them chat-invokable.
AI playground
Try it yourself.
Describe what you want back — the file, not the prompt engineering — and Agent P works out which of the 45 tools it takes.
Try these examples
Sign in and Agent P will run this
Real Agent P output Comes back as PNG · up to 4K
Powered by 45 Agent P tools
One conversation, end to end
45
tools, one chat
7
capability groups
4
output kinds
1
brief in, deliverable out
Capabilities
7 groups. Prompt to print file.
Every tool below is invoked the same way — you describe what you want, in the same chat, and Agent P chains whatever it takes to get there.
Create
14 toolsGenerate, edit, animate and combine assets.
Enhance & clean up
5 toolsFix, upscale, cut out and polish.
Trim, crop & resize
5 toolsTrim, reframe and resize for any platform.
Pull audio & frames out
3 toolsPull the audio, or grab a single frame.
Convert & export
6 toolsTurn your last mile into any format.
Audio & voice
10 toolsVoice-over, dubbing, stems and cleanup.
- One prompt. Multiple tools.
- Powered by 45 AI tools.
- No manual setup. Just create.
01 — 2 tools Max only
Plan the shootNew
The pre-production pass nobody else in agentic AI runs — lock the look before a single frame is generated.
- Board it, and lock the castNew Numbered panels with camera, lens and palette notes — and a character sheet that holds one face across every later shot.
- Turn footage into previsNew Clay-shaded depth render, optionally with pose skeletons.
02 — 14 tools Lite + Max
Create
Make the asset from nothing but a description — or from something you already have.
- Make an image Generate a brand-new image from a text description. Up to 4K.
- Edit or restyle an image Change style, add elements, composite or in-paint — up to 10 sources.
- Combine images Merge multiple images side-by-side, stacked or into a grid.
- Make a video from text Generate a clip from a prompt alone — no source media.
- Animate a still Turn a still into a moving clip, or tween an opening and closing frame.
- Keep a character across shotsNew Generate new video that holds the same faces from your references.
- Edit or restyle a video Anime, watercolour, comic or painterly — or recolour an object, change an attribute, swap a background. Original motion preserved.
- Add cinematic VFXNew Fire, smoke, explosions, muzzle flash, blood impacts, sparks, debris, rain, lightning, energy and screen shake — on footage you already have, same actors and camera.
- Render 2K videoNew A separate 2560×1440 path for when 720p will not do.
- Stitch clips together Join up to 12 clips with a crossfade or a hard cut.
- Continue a video Carry the final frame forward so the scene survives the cut.
- Make music Generate an original track from mood, genre and instruments.
- Generate sound effects Foley and ambience — or let it watch the footage and time them for you.
- Add or replace audio Swap a video’s soundtrack for a different track.
03 — 5 tools Max only
Enhance & clean up
Fix what the camera or the source file got wrong.
- Upscale an image Super-resolve a small or soft image to a sharper one.
- Upscale a video Raise a clip’s resolution — good for old footage.
- Remove an image background Cut the subject out and drop the background to transparent.
- Remove a video background Roto the subject out frame by frame. No green screen.
- Lip-sync a video Re-sync a talking head to a different audio track.
04 — 5 tools Max only
Trim, crop & resize
The edit-suite chores, done from a sentence instead of a timeline.
- Trim a video Keep the part you want and drop the rest.
- Crop or reframe Re-crop to 9:16 for Reels, 1:1 for feed, 16:9 for YouTube.
- Resize a video Scale to 480p through 4K, or to exact pixel dimensions.
- Change video speed Speed up or slow down, with pitch-corrected audio.
- Slice audio Cut an audio file down to the section you need.
05 — 3 tools Max only
Pull audio & frames out
Get the pieces out of a finished file so you can reuse them.
- Extract the audio Pull a clip’s soundtrack out as a standalone file.
- Remove the audio Strip the soundtrack and keep a silent video.
- Grab a still frame Freeze one moment as a thumbnail, poster or reference.
06 — 6 tools Max only
Convert & export
The last mile. This is the group nobody else in agentic AI ships.
- Convert a video format Transcode between MP4, MOV, MKV, AVI and WMV.
- Convert an image format PNG, JPG, WEBP, HEIC, AVIF — and rasterize an SVG.
- Video to animated GIF Two-pass palette encoding, so it is clean rather than muddy.
- Vectorize to SVG Trace a logo or flat graphic into genuinely scalable paths.
- Prepare for print Set an image to a real paper size and DPI.
- Convert subtitles Both directions between broadcast SCC and SRT, VTT and ASS.
07 — 10 tools
Audio & voice
A full voice room — script to voice-over, dubbing, dialogue, cloning, stems and cleanup.
- Transcribe to subtitles Turn spoken audio or video into a timed SRT or VTT.
- Text to speech Turn a script into a natural voice-over track.
- Clone a voiceNew Save a named voice from a reference clip and reuse it by @handle.
- Dub speech Re-voice dialogue in another language, from 30 preset voices.
- Multi-speaker dialogue Generate a scripted conversation with a voice per speaker.
- Split audio into stemsNew Separate vocals from the music-and-effects bed — the dubbing tool.
- Work out who spoke whenNew Speaker diarization: every segment labelled by speaker.
- Assemble an audio timeline Lay multiple pieces onto one timeline at exact timestamps.
- Normalize audio levels Two-pass EBU R128 loudness, to a real delivery standard.
- Denoise audio Strip hiss, hum and room noise out of a recording.
Lite + Max runs on the free tier. Max only tools need a Max session.
Free Lite sessions currently carry no credit cost — Max is charged at list price.
Vintage photo generator Upload one photo.
Upload one photo.
Get the 80s back.
Warm film colours, soft grain and classic studio backdrops — applied before you type a word. The styled photo opens in Agent P, so a different decade is the next thing you say, not the next tool you open.
- 01 Upload a photo Any photo works — selfie, portrait or group.
- 02 Get vintage styles Instant 80s, 90s, 2000s looks with real film feel.
- 03 Create & share Download or keep editing in Agent P.
Real vintage look Film grain & light leaks Multiple decades High-resolution results
Captions
Captions, from spoken word to SRT.
Point Agent P at a clip and get a timed caption file back — then transcode it to VTT, ASS or broadcast SCC in the same chat.
We hand you the caption file, not a burned-in frame. Your player or your editor does the overlay.
One conversation
- 01Transcribe the clip to a timed SRT
- 02Convert that SRT to VTT for the web player
- 03Convert it again to SCC for the broadcaster
- 04Crop the clip to 9:16 and export MP4
Transcription bills at 1 credit per minute of source. Subtitle conversion is free today.
Who it’s for
Built for the people who get paid to deliver files.
Print on demand
Design the artwork, cut it out, trace it to vector and prep it to the paper size and DPI your printer demands.
Delivers SVG · TIFF · transparent PNG
Social & Reels
Generate or animate the clip, stitch the cuts, reframe it to 9:16 and drop a music bed under it.
Delivers 9:16 MP4 · GIF
Paid ad delivery
Every ad platform wants a different aspect ratio and container. Reframe and transcode without leaving the chat.
Delivers MP4 · MOV · every ratio
Voice & localization
Voice a script, dub a talking head into another language, clean the audio up and re-sync the mouth to match.
Delivers audio · dubbed MP4 · SRT
Same models, from your code Agent P decides the chain.
Agent P decides the chain.
The API lets you write it.
Agent P is the conversation. Underneath it sits the same catalogue you can call directly — one endpoint per capability, one parameter to switch models.
377priced models on the API
47providers behind them
26endpoint pages
4capability families
Image 7 APIs
Text-to-image, image-to-image, editing, upscaling, restoration, reference-driven generation and background removal.
Video 10 APIs
Text-to-video, image-to-video, video-to-video, lipsync, plus reference- and audio-driven generation.
Audio 5 APIs
Speech, music, songs, voice clones and sound pulled from silent footage.
Everything else 5 APIs
3D assets, virtual try-on, LoRA training, consistent characters — and a free tier to build against.
Free to try, no card · keys and logs live in the API console
For developers One API.
One API.
Every AI capability.
Integrate 377 priced models through a single account. Build, scale and ship without holding an agreement with 47 separate providers.
- No separate contracts
- Unified billing
- SOC-2 & ISO 27001
curl -X POST \ "https://gateway.pixazo.ai/gpt-image-2-5-flare/v1/text-to-image" \ -H "Content-Type: application/json" \ -H "Cache-Control: no-cache" \ -H "Ocp-Apim-Subscription-Key: YOUR_API_KEY" \ --data-raw '{ "prompt": "A cinematic mountain landscape at sunset" }'import requestsr = requests.post( "https://gateway.pixazo.ai/gpt-image-2-5-flare/v1/text-to-image", headers={ "Content-Type": "application/json", "Ocp-Apim-Subscription-Key": "YOUR_API_KEY", }, json={"prompt": "A cinematic mountain landscape at sunset"},)job = r.json() # { request_id, status: "QUEUED", polling_url }print(job["polling_url"])const res = await fetch( "https://gateway.pixazo.ai/gpt-image-2-5-flare/v1/text-to-image", { method: "POST", headers: { "Content-Type": "application/json", "Ocp-Apim-Subscription-Key": process.env.PIXAZO_API_KEY, }, body: JSON.stringify({ prompt: "A cinematic mountain landscape at sunset" }), },);const job = await res.json(); // { request_id, status, polling_url }console.log(job.polling_url); 202 QUEUED Job accepted — poll polling_url for the result
Get started
Make one finished listing, free.
One brief. Every file it takes. Bring the idea and see what comes back — no card, no canvas to learn.
