Prism AI errors: transcribe, image, music, video

This page covers every message Prism AI can return when you transcribe, detect beats, or generate sound, music, an image or a video. First, read the message: a refused or failed generation is never charged, so nothing is lost.

How Prism AI is paid for

Prism AI is available on Pro and is paid from your Prism AI Balance, loaded in your dashboard. The balance is separate from your plan's MCP actions: an empty balance stops generation and nothing else. A request sent through your AI client also counts as one MCP action.

Balance and plan

Balance empty · ai-balance-exhausted

You see: "Prism AI Balance is {left}. This request needs about {est}. Load-to-use: buy Prism AI Balance to run this…"

Meaning. Your Prism AI balance is {balance}. This needs about {estimate}. Nothing was generated or charged.

Fix. Buy Prism AI balance in your dashboard, then run it again.

Your plan and every After Effects tool keep working. The panel's wording is "Out of Prism AI balance ($0.00 left). Top up to keep using AI tools — your plan and all other tools are unaffected."

Shows in: both.

Monthly limit reached · tier-cap-exceeded

You see: "You have used all {allowed} MCP actions included in your Prism {Plan} plan this month…"

Meaning. You have used this month's {allowance} on the {plan} plan. Nothing is broken and nothing was lost. Resets {date}.

Fix. Upgrade in your dashboard to continue now.

Shows in: your AI client.

Agent access off · ai-tool-disabled

You see: ""{label}" is turned off for agents in the Prism panel. Enable it under MCP → Agent AI Tools to let the agent use it. (It still works manually in the Tools panel.)"

Meaning. {Tool} is turned off for agents in the Prism panel. Nothing was generated or charged.

Fix. In the panel open MCP ▸ Agent AI Tools and turn {tool} on.

The switches govern only what an AI client may run; the Tools tab runs every tool by hand regardless.

Shows in: your AI client.

Before anything is generated

No composition open · panel-task

You see: "Open or create a composition first — no image was generated, so you were not charged."

Meaning. No composition is open to place the result in. Nothing was generated or charged.

Fix. Open or create a composition, then run it again.

Shows in: both.

Unknown AI action · input-validation

You see: "prism_ai action "{action}" generates from a description, and no prompt was given."

Meaning. That action does not exist. Nothing ran.

Fix. Use transcribe, detect_bpm, generate_image, generate_music, generate_sound_effect or generate_video.

Shows in: both.

Panel update needed · unknown-op

You see: "Unknown panel-task op: {op}"

Meaning. This panel version cannot run {op}. Nothing ran and nothing was charged.

Fix. Update the panel from your dashboard, then quit and reopen After Effects.

Shows in: both.

Server unreachable · server-unreachable

You see: "Failed to fetch"

Meaning. The panel could not reach Prism's server. Nothing ran and nothing was charged.

Fix. Check the panel shows Connected, then run it again.

Do not re-run a generation straight away; a second run is a second charge.

Shows in: both.

Input limits

Prism checks these before a paid call, so a refusal here costs nothing.

File too large · upload-too-large

You see: "That file is {MB} MB — Prism uploads up to 500 MB. Trim the media, or export just the audio, and try again."

Meaning. This file is {size} MB; Prism uploads up to {limit} MB. Nothing was sent or charged.

Fix. Trim the media, or export just the audio, then try again.

Shows in: both.

Transcription queue full

You see: "Prism is already transcribing as many files as it can at once."

Meaning. Prism is already transcribing as many files as it can. Nothing ran and nothing was charged.

Fix. Wait for the current transcription to finish, then send this one.

Shows in: both.

Resolution value invalid

You see: "For an image, resolution must be one of: 512, 1K, 2K, 4K." · "quality must be one of: auto, low, medium, high."

Meaning. The resolution is not an image size Prism accepts. Nothing was generated or charged.

Fix. Use 512, 1K, 2K or 4K for an image.

Shows in: your AI client.

Too many references

You see: "Too many reference images ({n}). Send at most 8." · "Reference image {i} is {mb} MB. The limit is 12 MB…" · "Reference image {i} must be an https:// URL or a data:image/…;base64 URI."

Meaning. {n} reference images were sent; the limit is 8. Nothing was generated or charged.

Fix. Send at most 8 reference images.

Shows in: your AI client.

Model not offered

You see: "Unknown {capability} model "{asked}". Prism runs one of: {names}. Omit model to let the server choose."

Meaning. {model} is not a model Prism runs for {capability}. Nothing was generated or charged.

Fix. Choose one of {names}, or leave the model unset.

Shows in: your AI client.

Video already generating · video-in-flight

You see: "A video is already generating. Wait for it to finish — starting another would generate and charge for a second clip."

Meaning. Nothing new started. A video is already generating; another would be charged as a second clip.

Fix. Wait for the current video to finish.

Shows in: the panel.

The generation provider rejected the request

These messages begin with the provider's name and an HTTP status, then a colon. The part after the colon is what to act on, quoted below. Nothing was charged.

Transcription rejected

You see: "All keywords must be less than 50 characters."

Meaning. The transcription was refused before it started. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Read the quoted reason, adjust the request, then run again.

Shows in: your AI client.

Transcription rejected

You see: "Some keyword contains invalid characters"

Fix. Read the quoted reason, adjust the request, then run again.

Shows in: your AI client.

Music request rejected

You see: "Invalid type of composition_plan used for model music_v2"

Meaning. The music was refused before it was made. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Describe the music in a prompt instead of a plan, then run again.

Shows in: your AI client.

Image request rejected

You see: "No provider for {model} supports the requested parameter(s): resolution "2K"… Accepted: 1K"

Meaning. The image was refused before it was made. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Use the accepted values quoted, or choose another model.

Shows in: your AI client.

Image request rejected

You see: "No provider for {model} supports the requested parameter(s): aspect_ratio "16:9"… Accepted: 1:1, 3:2, 2:3, auto"

Fix. Use the accepted values quoted, or choose another model.

Shows in: your AI client.

Video request rejected

You see: "Duration 5s is not supported for this model. Supported durations: 4, 6, 8s"

Meaning. The video was refused before it started. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Use the accepted values quoted, such as a supported duration.

Shows in: your AI client.

Timeouts

Transcription timed out

You see: "Transcription timed out after 10 minutes."

Meaning. The transcription did not return within 10 minutes. Nothing was charged.

Fix. Trim the audio to the part you need, then send that.

Shows in: both.

Video still generating · video-budget-exceeded

You see: "The video was still generating after 10 minutes. It may still finish upstream — try a shorter clip, a lower resolution, or a faster model."

Meaning. The video did not finish within 10 minutes. It may still complete, but nothing was charged.

Fix. Try a shorter clip, a lower resolution or a faster model.

Shows in: your AI client.

After the generation

Generated media goes offline

Generated files are written to a Prism folder inside your Documents; older builds used a temporary folder that macOS clears, so collect the files you keep into your project folder before closing it. Some H.264 files from video generators decode wrongly in After Effects, rendering black with no error: convert the clip to ProRes before importing. Both are in known issues.

Snapshot timed out · snapshot-timeout

You see: "No frames captured — The frame render did not finish within 30s. After Effects is usually waiting on something modal…"

Meaning. The frame did not render within 30 seconds; After Effects is usually waiting on a dialog. Nothing is left running.

Fix. Check the composition renders in the viewer at that time, then try again.

Shows in: your AI client.

Other Prism AI states

No layer selected · no-layer-selected

You see: "Select an audio layer in the composition. ‖ Select an audio layer in "${comp}" or pass layerName.…"

Meaning. No layer is selected, so Prism does not know which audio to use. Nothing ran.

Fix. Select the audio layer in the timeline, then run it again.

Several layers selected · too-many-selected

You see: "Select only ONE layer (got ${n}).…"

Meaning. More than one layer is selected, so Prism cannot tell which to use. Nothing ran.

Fix. Select just the one audio layer, then run it again.

Layer has no audio · no-audio-track

You see: "Selected layer has no audio track. ‖ "${name}" has no audio track to transcribe.…"

Meaning. The selected layer has no audio, so there is nothing to read. Nothing ran.

Fix. Select a layer with sound, such as a music or voice file.

Video file selected · video-not-audio

You see: "${name}" is a video file. Export audio first: Composition →…"

Meaning. The selected layer is a video, and Prism reads sound only from audio files. Nothing ran.

Fix. Export its audio (Composition ▸ Add to Render Queue ▸ Output Module ▸ Audio Only), then select it.

Layer has no file · no-source-file

You see: "Source has no file path (cannot read from disk). ‖ Selected layer has no source footage.…"

Meaning. The selected layer is not backed by a file on disk, so Prism cannot read its audio. Nothing ran.

Fix. Select a layer whose source is an audio file saved on disk.

Capability not configured

You see: "This capability is not configured on this server ({KEY_VAR}…"

Meaning. {Capability} is not set up on this server. Nothing ran and nothing was charged.

Fix. Email support@oneprism.io with the code.

No file uploaded

You see: "No file uploaded (multipart field "file" required).…"

Meaning. No audio or video file arrived with the request. Nothing ran.

Fix. Select an audio or video layer, or attach a file, then run again.

Audio too short

You see: "Decoded audio too short (< 1s)…"

Meaning. The audio is under one second, too short to find a tempo. Nothing was charged.

Fix. Use a clip of at least a few seconds.

Model not selectable

You see: "{capability} does not take a model choice — omit model.…"

Meaning. {capability} has no model to choose. Nothing was generated or charged.

Fix. Leave the model unset and run again.

Reference not text

You see: "Reference image {i} is not a string. Send an https:// URL or…"

Meaning. Reference image {i} is not text. Nothing was generated or charged.

Fix. Send each reference as an https:// URL or a data: URI.

Reference too large

You see: "Reference image {i} is {mb} MB. The limit is 12 MB — send a…"

Meaning. Reference image {i} is {size} MB; the limit is 12 MB. Nothing was generated or charged.

Fix. Send a smaller image, or link it by URL.

Reference link invalid

You see: "Reference image {i} must be an https:// URL or a data:image/…"

Meaning. Reference image {i} is not an https:// link or an inline image. Nothing was generated or charged.

Fix. Send each reference as an https:// URL or a data: URI.

No image returned

You see: "OpenRouter returned no image data.…"

Meaning. The image request succeeded but no image came back. Nothing was charged.

Fix. Run it again; if it repeats, choose another model.

Sound effect rejected

You see: "ElevenLabs error (HTTP {status}): {msg}…"

Meaning. The sound effect was refused before it was made. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Read the quoted reason, adjust the request, then run again.

Arrangement draft failed

You see: "Could not draft an arrangement: {raw}…"

Meaning. Prism could not draft the sections. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Describe the music more plainly, then run again.

Video job not tracked

You see: "OpenRouter accepted the job but returned no polling URL. Pay…"

Meaning. The video job was accepted but Prism cannot follow it. Nothing was charged.

Fix. Run it again; if it repeats, choose another model.

Video generation failed

You see: "The video model failed to generate: {reason}…"

Meaning. The video could not be generated. Nothing was charged. The generation provider rejected the request: {provider detail}

Fix. Simplify the prompt or shorten the clip, then run again.

No video returned

You see: "The job completed but returned no video URL.…"

Meaning. The video job finished but no video came back. Nothing was charged.

Fix. Run it again; if it repeats, choose another model.

Video download failed

You see: "Could not download the finished video (HTTP {n}).…"

Meaning. The video was generated but could not be fetched. Nothing was charged.

Fix. Run it again in a minute.

Video request timed out

You see: "OpenRouter did not respond in time while generating the vide…"

Meaning. No answer came back while generating the video. Nothing was charged.

Fix. Try a shorter clip, then run again.

Generation failed

You see: "(raw e.message)…"

Meaning. The generation stopped with an error. Nothing was charged.

Fix. Run it again; if it repeats, email support@oneprism.io with the code.

Audio decoding unavailable · ffmpeg-missing

You see: "FFmpeg not available: {msg}…"

Meaning. This server cannot decode audio right now. Nothing was charged.

Fix. Email support@oneprism.io with the code.

Audio not decodable · ffmpeg-missing

You see: "FFmpeg exit {code}: {stderr}…"

Meaning. Prism could not decode this audio for tempo detection. Nothing was charged.

Fix. Export the audio as WAV or MP3, then run again.

No tempo found

You see: "Beat tracking could not find a tempo in this audio (got: {te…"

Meaning. Prism could not find a steady beat in this audio. Nothing was charged.

Fix. Choose a section with a clear beat, or set the BPM by hand.

File type unsupported · upload-type-unsupported

You see: "Unsupported file type for transcription: {mt} (audio/video o…"

Meaning. {type} files cannot be used here; Prism takes {audio or video}. Nothing was charged.

Fix. Export the media as MP3, WAV or MP4, then try again.

Too many AI requests · ai-rate-limited

You see: "Too many requests, please try again later.…"

Meaning. More than 20 AI requests arrived in a minute. Nothing ran and nothing was charged.

Fix. Wait a minute, then run it again.

Frame not captured · frame-capture-failed

You see: "Could not capture the current frame: … Turn off "Animate cur…"

Meaning. The current frame could not be captured, so no video was started. Nothing was charged.

Fix. Turn off Animate current frame to generate from the prompt alone.

Common questions

If a message is not here, or the fix does not hold, email support@oneprism.io with the panel version (footer, "After Effects MCP v…") and the Copy diagnostic from the panel's Support section. On panel 1.5.1 the Support section may be empty; the version plus the error text as your AI client showed it is enough.

Related